PROPRIETARY INFORMATION INTERNAL USE ONLY THIS DOCUMENT AND THE DATA DISCLOSED HEREIN OR HEREWITH IS PROPRIETARY AND IS NOT TO BE REPRODUCED, USED OR DISCLOSED IN WHOLE OR IN PART TO ANYONE WITHOUT THE WRITTEN PERMISSION OF ORACLE. COPYRIGHT, © ORACLE 2022. ALL RIGHTS RESERVED. Title: R12.1 Reference Architecture B - Performance Test Plan (draft) Doc Number: TP10213 Revision #: 1.0 Policy Management Test Plan Reference Architecture B – Performance Test Plan for R12.1 Release 12.1 P. Piacentini DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 1 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 1 CHANGE HISTORY 4 2 REFERENCES 5 3 PURPOSE AND SCOPE 5 4 TERMINOLOGY 6 5 TOPOLOGIES 7 5.1 Reference B-1 Topology 5.1.1 Reference B-1 Performance metrics 7 7 5.2 Reference B-2 Topology 5.2.1 Reference B-2 Performance Metrics 8 8 5.3 Reference B-3 Topology 5.3.1 Wireless Outbound Roaming Section of Topology B-3 5.3.2 Wireless Inbound Roaming Section of Topology B-3 5.3.3 Reference B-3 Performance Metrics 9 9 10 10 5.4 Reference C-1 Topology 5.4.1 Reference C-1 Performance Metrics 10 11 6 SERVER AND SYSTEM CAPACITIES 12 7 CALL FLOWS 13 7.1 Call1 - LTE + Subscriber Quota(2 PDN Connections) 14 7.2 Call2 - Subscriber Quota(2 PDN Connections) Allot Variation 15 7.3 Call3 - LTE No Quota 2 PDN connection +Sy 16 7.4 Call4 - GPRS(2 PDN Connections) + Pooled Quota + Regular Subscriber Quota 17 7.5 Call5 - GPRS(2 PDN Connections) with secondary PDP + Secondary Gx Subscriber Quota 18 7.6 Call6 - IMS VoLTE Call (VoLTE calls will contain both IMSI and E164 Sub ID) 19 7.7 Call7 - Application Monitoring on Sd 20 7.8 Call8 - LTE + Quota on Sd 21 7.9 Variation1 - OCS Triggered Throttling 21 7.10 Variation2 - Sh Triggered Throttling 22 DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 2 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 7.11 VoLTE plus S9 OutBound Roaming 22 7.12 VoLTE plus S9 InBound Roaming 22 8 GENERAL CONFIGURATION SETUP 22 9 EXAMPLE POLICIES 23 10 GENERAL GOALS OF THE TESTS 23 10.1 Latency Goals 24 10.2 Stability Goals 24 10 TEST SUITE : TP010213_REFERENCE ARCHITECTURE B PERFORMANCE TEST PLAN 25 10.1 Test Suite : Reference Architecture B-1 25 10.1.1 Test Case CAM-94636: Mixed call flow at 80% using IPv4 in call flow 25 10.1.2 Test Case CAM-94637: Mixed call flow at 100% using IPv4 in call flow 26 10.1.3 Test Case CAM-94638: Mixed call flow at 120% using IPv4 in call flow 27 10.1.4 Test Case CAM-94639: Mixed call flow at 150% using IPv4 in call flow 28 10.1.5 Test Case CAM-94640: Mixed call flow at 200% using IPv4 in call flow 29 10.1.6 Test Case CAM-94641: Mixed call flow at 80% using IPv6 in call flow 30 10.1.7 Test Case CAM-94642: Mixed call flow at 100% using IPv6 in call flow 31 10.1.8 Test Case CAM-94643: Mixed call flow at 120% using IPv6 in call flow 31 10.1.9 Test Case CAM-94644: Mixed call flow at 150% using IPv6 in call flow 32 10.1.10 Test Case CAM-94645: Mixed call flow at 200% using IPv6 in call flow 33 10.1.11 Test Case CAM-95599: Mixed call flow at 100% using Dual-Stack IP in call flow 34 10.1.12 Test Case CAM-94647: Mixed call flow at 100% using IPv4 in call flow in all IPv6 topology 35 10.1.13 Test Case CAM-95602: Mixed call flow at 100% using IPv4 in call flow in mixed IPv4/IPv6 topology 36 10.1.14 Test Case CAM-97266: Mixed call flow at 100% using IPv4 in call flow with Topology Hiding 37 10.1.15 Test Case CAM-94648: Mixed call flow at 100% using IPv4 in call flow using SCTP transport 38 10.1.16 Test Case CAM-95432: BL460 MPE with 64GB RAM can Support 6.5 Million Concurrent Sessions in B-1 Topology 39 10.1.17 Test Case CAM-95433: BL460 MRA with 64GB RAM can support up to 52,000 Transactions/second in B-1 Topology 40 10.1.18 Test Case CAM-95619: BL460 MRA with 64GB RAM can support up to 20M bindings in B-1 Topology 41 10.1.19 Test Case CAM-94646: Longevity Test - Mixed call flow at 100% using IPv4 in call flow 42 10.1.20 Test Case CAM-95439: CMP Can Support Managing up to 50 MPEs/MRA Geo-Redundant Clusters 43 10.1.21 Test Case CAM-95437: BL460 MRA with 64GB RAM can support up to 8k Client Connections 44 10.1.22 Test Case CAM-97374: MRA will support 16 Diameter connections from a DRA 46 10.1.23 Test Case CAM-98672: Baseline Step Test - BL460-G8 MPE 46 10.1.24 Test Case CAM-98673: Baseline Step Test - BL460-G8 MRA 47 10.1.25 Test Case CAM-99004: Baseline Step Test - BL460-G9 MPE 48 10.1.26 Test Case CAM-99005: Baseline Step Test - BL460-G9 MRA 49 10.1.27 Test Case CAM-98677: Mixed call flow at 100% loads with over 1k Policies Deployed 50 10.1.28 Test Case CAM-98702: 2-Site MRA at 100% load in Legacy mode 51 10.1.29 Test Case CAM-98703: 2-Site MRA at 100% load in Algov1 mode 51 DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 3 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 10.2 Test Suite : Reference Architecture B-2 52 10.2.1 Test Case CAM-95618: BL460 MRA with 64GB RAM can support up to 52,000 Transactions/second in B-2 Topology 52 10.2.2 Test Case CAM-95435: BL460 MRA with 64GB RAM can support up to 20M bindings in B-2 Topology 54 10.3 Test Suite : Reference Architecture B-3 10.3.1 Test Case CAM-94649: Mixed call flow at 80% using IPv4 in call flow 10.3.2 Test Case CAM-94650: Mixed call flow at 100% using IPv4 in call flow 10.3.3 Test Case CAM-94651: Mixed call flow at 120% using IPv4 in call flow 10.3.4 Test Case CAM-94652: Mixed call flow at 150% using IPv4 in call flow 10.3.5 Test Case CAM-94653: Mixed call flow at 200% using IPv4 in call flow 10.3.6 Test Case CAM-94654: Longevity Test - Mixed call flow at 100% using IPv4 in call flow 10.3.7 Test Case CAM-94655: Mixed call flow at 100% using IPv4 in call flow using SCTP transport 10.3.8 Test Case CAM-95434: Oracle X5-2 MRA can support up to 50,000 Transactions/second 10.3.9 Test Case CAM-95431: Oracle X5-2 MPE can Support 10 Million Concurrent Sessions 10.3.10 Test Case CAM-95440: Oracle X5-2 MRA can support up to 40M bindings 10.3.11 Test Case CAM-95438: Oracle X5-2 MRA can support up to 8k Client Connections 10.3.12 Test Case CAM-97382: VoLTE-Plus-Inbound-S9 Roaming call flow at 100% 10.3.13 Test Case CAM-97323: VoLTE-Plus-Outbound-S9 Roaming call flow at 100% 55 55 56 56 57 58 59 60 60 62 63 64 65 66 10.4 Test Suite : Reference Architecture C-1 10.4.1 CAM-94630:RADIUS with Sd and SPR Profile Lookup at 80% load 10.4.2 CAM-94631:RADIUS with Sd and SPR Profile Lookup at 100% load 10.4.3 Test Case CAM-94632: RADIUS with Sd and SPR Profile Lookup at 120% load 10.4.4 Test Case CAM-94633: RADIUS with Sd and SPR Profile Lookup at 150% load 10.4.5 Test Case CAM-94634: RADIUS with Sd and SPR Profile Lookup at 200% load 10.4.6 Test Case CAM-94635: RADIUS with Sd and SPR Profile Lookup Longevity Test 68 68 69 70 71 72 73 10.5 Test Suite : Reference Architecture B Impairment tests 74 10.5.1 Test Case CAM-95400: BL460 C-Class Active MPE Failover to Standby MPE at 100% Load 74 10.5.2 Test Case CAM-99006: BL460-G9 C-Class Active MPE Failover to Standby MPE at 100% Load 76 10.5.3 Test Case CAM-95401: Oracle X5-2 Active MPE Failover to Standby MPE at 100% Load 77 10.5.4 Test Case CAM-95402: BL460 C-Class Geo-Redundant Site1 MPE Failover to Site2 MPE at 100% Load 77 10.5.5 Test Case CAM-95403: Oracle X5-2 Geo-Redundant Site1 MPE Failover to Site2 MPE at 100% Load 78 10.5.6 Test Case CAM-95404: BL460 C-Class Active MRA Failover to Standby MRA at 100% Load 79 10.5.7 Test Case CAM-99007: BL460-G9 C-Class Active MRA Failover to Standby MRA at 100% Load 80 10.5.8 Test Case CAM-95405: Oracle X5-2 MRA Failover to Standby MRA at 100% Load 81 10.5.9 Test Case CAM-95406: BL460 C-Class Geo-Redundant primary site MRA Failover to secondary site MRA at 100% Load 82 10.5.10 Test Case CAM-95407: Oracle X5-2 Geo-Redundant primary site MRA Failover to secondary site MRA at 100% 83 10.5.11 Test Case CAM-95408: Surge of Session Creation Requests 83 10.5.12 Test Case CAM-98453: 100% Load to MRA with 80ms Latency Added Between MRA and MPE 84 10.5.13 Test Case CAM-98454: 100% Load to MRA with 80ms Latency Added Between MPEs and SPR 85 10.5.14 Test Case CAM-99096: 100% Load to 2-site MRAs with dropped DRMA messages between MRAs 87 Appendix A. Review Summaries 89 1 CHANGE HISTORY DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 4 of 94 PROPRIETARY INFORMATION Date 4/30/2015 5/12/2015 5/28/2015 6/11/2015 6/15/2015 7/13/2015 2 Rev # 0.1 0.2 0.3 0.4 0.5 1.0 INTERNAL USE ONLY Author P. Piacentini P. Piacentini P. Piacentini P. Piacentini P. Piacentini P. Piacentini Approved No No No No No Yes REFERENCES Title [1] Product Function Specification (PFS) 12.1 [2] Reference Architecture for Release 12.1 [3] [4] [5] 3 Revision Description 1st draft Changes based on online review comments Changes based on in-person review comments Added some test cases for BL460-G9 Added test case where messages dropped between MRAs Document approved Number PF006194 FE007538 revision 1.3 1.1 Author C. Berlutti C. Berlutti Date 04/15/2015 03/24/2015 PURPOSE AND SCOPE This test plan is based on the configurations and call flows as described in FE007538 - Reference Architecture for Release 12.1. The tests contained within are intended to verify the requirements that are described in that document. DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 5 of 94 PROPRIETARY INFORMATION 4 INTERNAL USE ONLY TERMINOLOGY Binding A binding between a subscriber identifier (e.g. IMSI, MSISDN, IP Address) and an MPE. The MRA maintains bindings, and there is one binding per subscriber even if the subscriber has multiple active sessions. BRAS/BNG A BRAS or BNG is where the access session is terminated. It also serves as policy enforcement point for fixed line networks. In theory, a BRAS supports PPP sessions across ATM, while BNG supports native IP over Ethernet (using DHCP extensions to define each subscriber) across an Ethernet WAN. In reality, many current devices support both network types, so the terms are often used interchangeably. The terms BRAS and BNG are used interchangeably in this document. Other terms describing this device include edge router and services router. Policy Solution The total of all Policy Components (CMPs, MRAs, MPEs) and Policy Systems (see below) that comprise the policy deployment for a network. A Policy Solution may contain one or more Policy Systems. Policy System All of the Policy Components under the control of a single NW-CMP instance. A single NW-CMP instance with multiple S-CMP may host multiple MRAs and multiple MPEs behind the MRAs. A Policy System will generally consist of at least two Sites, since a NW-CMP instance generally hosts components at two mated, geo-diverse Sites. PCRF Segment A partition of the Policy function within an operator’s network • Policy segments may be defined by functionality, scale, geography, PCRF vendor or other operator driven partitioning motive • Intra-segment Policy session correlation may or may not be provided by the PCRF • Inter-segment Policy Session correlation is NOT provided by the PCRF For the Oracle PCRF, a Policy Segment is a collection of one or more MPE clusters, fronted by one or more PFE Clusters and managed by a single CMP Cluster. An Oracle Policy Segment: • May span multiple PCRF Nodes • May span multiple PCRF Sites • Provides intra-segment Policy session correlation Session A Diameter session between the MPE and an external device (e.g. a Gx, Gxa, Gx-Lite, Rx and Sy session), Sh, as well as non-diameter sessions, are excluded. Subscribers can (and typically do) maintain multiple sessions at any given time. Site All Policy Components physically hosted at a single location. Transaction An incoming or outgoing message (such as a Diameter CCR-I or an RAR) and all the associated internal product processing until the subsequent message is sent/received. Note that the subsequent message need not be of the same type as the initial message. For example: - Gx CCR-u incoming Diameter message, internal processing, Gx CCA Diameter response message = 1 transaction - Gx CCR-i incoming Diameter message, internal processing, Sh UDR outgoing Diameter HSS query message, Sh UDA Diameter response message, internal processing, Gx CCA Diameter response message = 2 Transactions DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 6 of 94 PROPRIETARY INFORMATION 5 INTERNAL USE ONLY DRMA transactions between MRAs in multi-site deployments, and transactions between MRA and MPEs, are not included in this rated capacity. All transactions between the MPE and GWs, AFs, MRAs, SPRs, LDAP Servers, OCSs, and SMSC Servers are included in this capacity. TOPOLOGIES 5.1 REFERENCE B-1 TOPOLOGY Reference B-1 topology consists of all Geo-Redundant MRA and MPE clusters. There will be two MRA sites. The second MRA site will most likely not be a full chassis of blades, depending on hardware availability. All C-Class blades are BL460 G8. The RMS is DL380. Subscriber authentication will be performed by LDAP and/or SPR lookup. Quota and Entity State variables will be managed and stored by the SPR. Both Legacy SPR and R10.0 OCUDR will be used. TDF and OCS functions will be simulated by seagull servers. Although the DUT will be the MRA and MPEs at Site-1, for every test there will be a small amount of traffic routed between Site-1 and Site-2 5.1.1 REFERENCE B-1 PERFORMANCE METRICS Component Performance Target (per instance) DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. System TPS capability Title: System TPS requirement Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 7 of 94 PROPRIETARY INFORMATION 5.2 MRA Per PCRF GeoRedundant Segment(Statefull, Dynamic)- (1) MRA cluster 52,000 TPS 40M Bindings MPE Per PCRF GeoRedundant Segment- (7) MPE Clusters 6,500 TPS for Gen 8 6,500 TPS for Gen 9 15M Concurrent Sessions Gen 9 15M Concurrent Session/Gen 8 INTERNAL USE ONLY 52,000 (100% rated, no failures) 41,600 TPS@80% HP Gen 9 & HP Gen 8 BL460 Server Clusters 45,500 TPS@100% HP Gen 9 & HP Gen 8 BL460 Server Clusters 105M Concurrent Sessions@100% HP Gen 9 & HP Gen 8 BL460 Server Clusters HP Gen 9 & HP Gen 8 BL460 Server Clusters 36,400 TPS@80% HP Gen 9 & HP Gen 8 BL460 Server Clusters 84M Concurrent Sessions@80% HP Gen 9 & HP Gen 8 BL460 Server Clusters REFERENCE B-2 TOPOLOGY Topology B-2 consists of a single fully loaded chassis, consisting of an MRA cluster and seven MPE clusters. The chassis is managed by a rack-mount CMP cluster. All C-Class blades are BL460 G8. The RMS is DL380. Just a minimal set of tests will be run in this configuration, since it’s a subset of topology B-1 and will inherently be covered by tests run on that topology. 5.2.1 REFERENCE B-2 PERFORMANCE METRICS Component (1) MRA Active/Standby Performance Target (per instance) 52,000 TPS 10K PCRF Client connections 40M Bindings System TPS capability System TPS requirement 52,000 (100% rated, no failures) 41,600 TPS@80% 6,500 TPS for Gen 8 6,500 TPS for Gen 9 45,500 TPS@100% HP Gen 9 & HP Gen 8 HP Gen 9 & HP Gen 8 BL460 Server Clusters HP Gen 9 & HP Gen 8 BL460 Server Clusters Cluster (7) MPE Active/Standby DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: 36,400 TPS@80% HP Gen 9 & HP Gen 8 Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 8 of 94 PROPRIETARY INFORMATION Clusters 5.3 INTERNAL USE ONLY 15M Concurrent Sessions Gen 9 15M Concurrent Session/Gen 8 BL460 Server Clusters 105M Concurrent Sessions@100% HP Gen 9 & HP Gen 8 BL460 Server Clusters BL460 Server Clusters 84M Concurrent Sessions@80% HP Gen 9 & HP Gen 8 BL460 Server Clusters REFERENCE B-3 TOPOLOGY Topology B-3 consists of all rack-mount servers. In this case the servers will be Oracle X5-2 servers. The setup consists of an RMS CMP cluster, and RMS MRA cluster, and 1 or more RMS MPE clusters. 5.3.1 WIRELESS OUTBOUND ROAMING SECTION OF TOPOLOGY B-3 DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 9 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 5.3.2 WIRELESS INBOUND ROAMING SECTION OF TOPOLOGY B-3 5.3.3 REFERENCE B-3 PERFORMANCE METRICS Component (1) MRA Performance Target (per instance) System TPS capability System TPS requirement 52,000 TPS 40M Bindings 52,000 (100% rated, no failures) 41,600 TPS@80% 6,500 TPS for Gen 8 15M Concurrent Sessions 13,000 TPS@100% 30M Concurrent Sessions@100% 10,400 TPS@80% 44M Concurrent Sessions@80% Active/Standby Cluster (2) MPE Active/Standby Clusters 5.4 REFERENCE C-1 TOPOLOGY Topology C-3 consists of all rack-mount servers. The same Oracle X5-2 servers that were used in topology B-3 will be used in this setup. The setup consists of an RMS CMP cluster, and RMS MRA cluster, and 1 or more RMS DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 10 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY MPE clusters. Topology C-3 will focus on the Fixed Mobile Convergence use case, with a mix of RADIUS and Diameter interfaces. 5.4.1 REFERENCE C-1 PERFORMANCE METRICS Component (1) MRA Performance Target (per instance) System TPS capability System TPS requirement 52,000 TPS 40M Bindings 52,000 (100% rated, no failures) 41,600 TPS@80% 6,500 TPS 15M Concurrent Sessions 13,000 TPS@100% 30M Concurrent Sessions@100% 10,400 TPS@80% 44M Concurrent Sessions@80% Active/Standby Cluster (2) MPE Active/Standby Clusters DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 11 of 94 PROPRIETARY INFORMATION 6 INTERNAL USE ONLY SERVER AND SYSTEM CAPACITIES The following table described the performance capacities of the various hardware combinations that are supported in Release 12.1. Hardware TPS Per-MPE Capacity (100% Rated Capacity) Conc. Sub. Sessions Bindings 6.5M N/A TPS Per-MRA Capacity (100% Rated Capacity) Conc. Sessions Sub. Bindings BL460 G6 4500 25,200 N/A 20M DL360/380 G6 4500 6.5M N/A 25,200 N/A 20M BL460 G8 (64G RAM) BL460 G8+ (64G RAM) BL460 G8 (256G RAM DL 380 G8 (64G RAM) DL 380 G8 + (64G RAM) BL460 G9 (256G RAM) DL 380 G9 (128G RAM) Oracle Netra X32 RMS (128G RAM) Oracle X5-2 RMS (256G RAM) 6000 6.5M N/A 50k N/A 20M* 6000 6.5M N/A 50k N/A 20M* 6500 15M N/A 52K NA 40M 6000 6.5M N/A 50k N/A 20M* 6000 6.5M N/A 50k N/A 20M* 6500 15 M N/A 52k N/A 40M 6000 10 M N/A 50k N/A 40M 6000 10M N/A 50k N/A 40M 6500 15M N/A 52k N/A 40M DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 12 of 94 PROPRIETARY INFORMATION 7 INTERNAL USE ONLY CALL FLOWS Each test will use a mix of call flows. This section descibes each of the call flows used in detail. The percentage of each call type that will be attempted in the mixed calls traffic model is shown in the table below. Percentage of Total Traffic Call Flow Name LTE + Subscriber Quota(2 PDN Connections) 5 LTE + Subscriber Quota(2 PDN Connections) Allot Variation 5 LTE No Quota (2 PDN connections) +Sy 10 GPRS(2 PDN Connections) + Pooled Quota + Regular Subscriber Quota 10 GPRS(2 PDN Connections) with secondary PDP + Secondary Gx Subscriber Quota) 10 IMS VoLTE Call + Sy OCS Triggered Throttling 25 5 Sh Triggered Throttling Application Monitoring on Sd LTE + Quota on Sd VoLTE + S9 Outbound Roaming VoLTE + S9 InBound Roaming Total Percentage DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. 5 5 10 5 5 100 Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 13 of 94 PROPRIETARY INFORMATION 7.1 INTERNAL USE ONLY Call1 - LTE + Subscriber Quota(2 PDN Connections) DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 14 of 94 PROPRIETARY INFORMATION 7.2 INTERNAL USE ONLY CALL2 - SUBSCRIBER QUOTA(2 PDN CONNECTIONS) ALLOT VARIATION DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 15 of 94 PROPRIETARY INFORMATION 7.3 INTERNAL USE ONLY CALL3 - LTE NO QUOTA 2 PDN CONNECTION +SY DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 16 of 94 PROPRIETARY INFORMATION 7.4 INTERNAL USE ONLY CALL4 - GPRS(2 PDN CONNECTIONS) + POOLED QUOTA + REGULAR SUBSCRIBER QUOTA DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 17 of 94 PROPRIETARY INFORMATION 7.5 INTERNAL USE ONLY CALL5 - GPRS(2 PDN CONNECTIONS) WITH SECONDARY PDP + SECONDARY GX SUBSCRIBER QUOTA DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 18 of 94 PROPRIETARY INFORMATION 7.6 INTERNAL USE ONLY CALL6 - IMS VOLTE CALL (VOLTE CALLS WILL CONTAIN BOTH IMSI AND E164 SUB ID) DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 19 of 94 PROPRIETARY INFORMATION 7.7 INTERNAL USE ONLY CALL7 - APPLICATION MONITORING ON SD DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 20 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 7.8 CALL8 - LTE + QUOTA ON SD 7.9 VARIATION1 - OCS TRIGGERED THROTTLING A small percentage of the calls that use an OCS will have an SNR from the OCS trigger an RAR to the Gx interface to the gateway. DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 21 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 7.10 VARIATION2 - SH TRIGGERED THROTTLING A small percentage of the calls will have a PNR sent from the SPR, which will triggers an RAR on the Gx interface to the gateway, 7.11 VOLTE PLUS S9 OUTBOUND ROAMING This call flow will use the same script and policies used in “Call6 – IMS VoLTE Call”. The Policy Server configuration will be changed to support the S9 configuration and topology 7.12 VOLTE PLUS S9 INBOUND ROAMING This call flow will use the same script and policies used in “Call6 – IMS VoLTE Call”. The Policy Server configuration will be changed to support the S9 configuration and topology 8 GENERAL CONFIGURATION SETUP The MRAs are configured in stateful mode, as backups of each other All C-Class blades will have mezzanine cards installed, and the traffic will be separated so that OAM traffic uses bond0, and signaling traffic will use bond1 All MPEs will be configured with “Validate user” enabled All MPEs will have Primary and Secondary HSS Data Sources configured, using HSS profilev4 Subscription will be enabled on the HSS Data Sources A subscription-Id of E164 will be used MRA and MPEs indexed by E164 and IP, MRA also indexed by Session-ID, and primary indexing on the MRA is E164 Prior to running the performance tests, each of the MPEs will be preloaded with 1M sessions. Prior to running the tests, all statistics will be cleared Session cleanup will be disabled so that these sessions stay intact during the tests. Each test will run for 2 hours, and memory, CPU, latency, and other KPIs will be collected during the test The CMP will be configured to send SNMP traps to a remote trap receiver The CMP will be configured to forward syslog messages to a remote syslog server DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 22 of 94 PROPRIETARY INFORMATION 9 INTERNAL USE ONLY EXAMPLE POLICIES Many of the scripts and policies that will be needed to support these test cases have not yet been developed, therefore they cannot be included in this test plan. Once the testing gets underway, it is assumed that they will be accessable to anyone that wouild be interested in looking at them or trying to use them. Unless it proves too difficult to implement, the current plan is to use the APN in the policy conditions to trigger the correct policies based on the expected call flow. Also, each call flow script will use a unique set of Subscription-Ids so that the calls can be differentiated. Below is an example of the policies that could be used for Call #1: PERF_call1_int_init_save_state where the request is creating a new session And where the session is an enforcement session And where the device type is not DPI And where the APN matches one of call1_int And where the network element name matches one of ggsn-1,ggsn-2,ggsn-3,ggsn-4,ggsn-5,ggsn-6,ggsn-7 set the subscriber state variable call1_session_start to now using configured local time and save always set the subscriber state variable call1_reset to now + 1 minutes rounded down with same granularity using CONFIGURED LOCAL TIME and save always continue processing message PERF_call1_int_grant_subscriber_quota where the request is creating a new session,modifying an existing session And where the session is an enforcement session And where the APN matches one of call1_int grant total volume to 100 percent used for PERF_call1_subscriber using call1_int set session revalidation time to 60 seconds continue processing message PERF_call1_int_grant_subscriber_quota_on_update where the request is modifying an existing session And where the session is an enforcement session And where the APN matches one of call1_int grant total volume to 100 percent used for PERF_call1_subscriber using call1_int continue processing message PERF_call1_int_term_reset_quota_save_state where the request is terminating an existing session And where the session is an enforcement session And where the device type is not DPI And where the APN matches one of call1_int And where the network element name matches one of ggsn-1,ggsn-2,ggsn-3,ggsn-4,ggsn-5,ggsn-6,ggsn-7 set the subscriber state variable call1_session_stop to now using CONFIGURED LOCAL TIME and save always reset all plan usage continue processing message 10 GENERAL GOALS OF THE TESTS DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 23 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 10.1 LATENCY GOALS For each test that sends traffic at 80% and 100% of the rated capacity of the DUT, the following latency values are expected to be met: 1. 92% of the calls are expected to be achieved with a latency of under 100 ms 2. 99.9% of the calls are expected to be achieved with a latency of under 200 ms 3. The transaction response time of the Policy system with two sites for 92% of all transactions shall not exceed 167 + (2 * LL) ms; and 4. The transaction response time of the Policy system with two sites for 99.9% of all transactions shall not exceed 250 + (2 * LL) ms, where LL is the latency of the link between the two systems 5. The latency from the SPR is expected and assumed to be 25ms or less as part of the transaction. 6. The latency from the OCS simulator is expected and assumed to be 20ms or less as part of the transaction 7. The latency from the PCEF and AF simulator is expected and assumed to be 20ms or less as part of the transaction. 10.2 STABILITY GOALS For each test case that sends traffic at 120% or more of the rated capacity of the DUT, the following expectations are to be met: 1. There will be no unexpected exceptions seen in the logs of the MRA or MPEs 2. There will be no unexpected failure or failover of the MRA or MPEs 3. Some load shedding at increasingly higher percentage is expected as the traffic is increased beyond the rated capacity a. 4. The systems are expected to return no nor,normal serviceabbilty when the traffic is lowered to fall within it’s rated capacity The systems are expected to recover after each of the impariment tests are run DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 24 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 10 TEST SUITE : TP010213_REFERENCE ARCHITECTURE B PERFORMANCE TEST PLAN This test plan is for R12.1. 10.1 TEST SUITE : REFERENCE ARCHITECTURE B-1 Reference Architecture B-1 Reference Architecture B-1 consists of an RMS CMP, along with a full C-Class chassis, containing 1 MRA blade pair, and 7 MPE blade pairs. It also includes a second C-Class chassis half populated with blades, including an MRA, and 7 MPEs, which are configured in the topology as a second site to the first chassis, providing Geo-Redundancy. A second backup MRA site will also be set up, and for each test a small amount of traffic will be routed between the MRAs. · · · · · HP Gen8 servers will be used for these tests. There are a few Gen8Plus blades in the setup as well The MRAs are configured in stateful mode. All C-Class blades will have mezzanine cards installed, and the traffic will be separated so that OAM traffic uses bond0, and signaling traffic will use bond1 Prior to running the performance tests, each of the MPEs will be preloaded with 1M sessions. Session cleanup will be disabled on most of these tests, so that the preloaded sessions stay intact during the tests. Session cleanup will be enabled for the longevity tests 10.1.1 TEST CASE CAM-94636: MIXED CALL FLOW AT 80% USING IPV4 IN CALL FLOW Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 80% of the capacity of the MPEs. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 80% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. Expected Results: DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 25 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-252: R-19526026-252 The general traffic model described in [R-19526026230] shall be repeated on hardware R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t R-19526026-600: R-19526026-600 The following SPR/OCUDR Configurations and as shown in Table 15 shall be verified usi Keywords: Performance ANCHORAGE 10.1.2 TEST CASE CAM-94637: MIXED CALL FLOW AT 100% USING IPV4 IN CALL FLOW Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 100% of the capacity of the MPEs. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 100% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 26 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-252: R-19526026-252 The general traffic model described in [R-19526026230] shall be repeated on hardware R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t R-19526026-600: R-19526026-600 The following SPR/OCUDR Configurations and as shown in Table 15 shall be verified usi Keywords: Performance ANCHORAGE 10.1.3 TEST CASE CAM-94638: MIXED CALL FLOW AT 120% USING IPV4 IN CALL FLOW Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 120% of the capacity of the MPEs. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 120% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. Expected Results: The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 27 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY testing MPE and MRA shall include ad R-19526026-252: R-19526026-252 The general traffic model described in [R-19526026230] shall be repeated on hardware R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t Keywords: 10.1.4 Performance ANCHORAGE TEST CASE CAM-94639: MIXED CALL FLOW AT 150% USING IPV4 IN CALL FLOW Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 150% of the capacity of the MPEs. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 150% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. 6. Verify that if load shedding kicks in, the correct levels and discards are used Expected Results: The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-252: R-19526026-252 The general traffic model described in [R-19526026- DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 28 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 230] shall be repeated on hardware R-19526026-260: R-19526026-260 The mixed call model defined in Section 2.3.14 shall be repeated on hardware configu R-19526026-270: R-19526026-270 The mixed call model defined in Section 2.3.14 shall be repeated on hardware configu R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t Keywords: 10.1.5 Performance ANCHORAGE TEST CASE CAM-94640: MIXED CALL FLOW AT 200% USING IPV4 IN CALL FLOW Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 200% of the capacity of the MPEs. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 200% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. 6. Verify that if load shedding kicks in, the correct levels and discards are used Expected Results: The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-252: R-19526026-252 The general traffic model described in [R-19526026- DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 29 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 230] shall be repeated on hardware R-19526026-260: R-19526026-260 The mixed call model defined in Section 2.3.14 shall be repeated on hardware configu R-19526026-270: R-19526026-270 The mixed call model defined in Section 2.3.14 shall be repeated on hardware configu R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t Keywords: 10.1.6 Performance ANCHORAGE TEST CASE CAM-94641: MIXED CALL FLOW AT 80% USING IPV6 IN CALL FLOW Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 80% of the capacity of the MPEs. The format of the IP address in the calls will be IPv6 Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 80% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. Expected Results: The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t Keywords: Performance DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 30 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY ANCHORAGE 10.1.7 TEST CASE CAM-94642: MIXED CALL FLOW AT 100% USING IPV6 IN CALL FLOW Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 80% of the capacity of the MPEs. The format of the IP address in the calls will be IPv6 Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 100% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. Expected Results: The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t Keywords: Performance ANCHORAGE 10.1.8 Author: TEST CASE CAM-94643: MIXED CALL FLOW AT 120% USING IPV6 IN CALL FLOW ppiacentini DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 31 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 80% of the capacity of the MPEs. The format of the IP address in the calls will be IPv6 Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 120% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. Expected Results: The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t Keywords: Performance ANCHORAGE 10.1.9 TEST CASE CAM-94644: MIXED CALL FLOW AT 150% USING IPV6 IN CALL FLOW Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 80% of the capacity of the MPEs. The format of the IP address in the calls will be IPv6 Steps: DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 32 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 150% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. 6. Verify that if load shedding kicks in, the correct levels and discards are used Expected Results: The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-260: R-19526026-260 The mixed call model defined in Section 2.3.14 shall be repeated on hardware configu R-19526026-270: R-19526026-270 The mixed call model defined in Section 2.3.14 shall be repeated on hardware configu R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t Keywords: Performance ANCHORAGE 10.1.10 TEST CASE CAM-94645: MIXED CALL FLOW AT 200% USING IPV6 IN CALL FLOW Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 80% of the capacity of the MPEs. The format of the IP address in the calls will be IPv6 Steps: DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 33 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 200% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. 6. Verify that if load shedding kicks in, the correct levels and discards are used Expected Results: The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-260: R-19526026-260 The mixed call model defined in Section 2.3.14 shall be repeated on hardware configu R-19526026-270: R-19526026-270 The mixed call model defined in Section 2.3.14 shall be repeated on hardware configu R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t Keywords: Performance ANCHORAGE 10.1.11 TEST CASE CAM-95599: MIXED CALL FLOW AT 100% USING DUAL-STACK IP IN CALL FLOW Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 100% of the capacity of the MPEs. The UE address in the Gx messages will contain both IPv4 and IPv6 IP addresses Steps: DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 34 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 100% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. The MRA and MPE should process 100% of the calls generated by the traffic generator, using dualstack IP addressing. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-251: R-19526026-251 The general traffic model described in [R-19526026230] shall be repeated on hardware R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t Keywords: Performance ANCHORAGE 10.1.12 TEST CASE CAM-94647: MIXED CALL FLOW AT 100% USING IPV4 IN CALL FLOW IN ALL IPV6 TOPOLOGY Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 100% of the capacity of the MPEs. The topology will be using all IPv6 addresses for OAM, signaling, and replication. Steps: DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 35 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 100% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-250: R-19526026-250 The general traffic model described in [R-19526026230] shall be repeated on hardware R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t Keywords: Performance ANCHORAGE 10.1.13 TEST CASE CAM-95602: MIXED CALL FLOW AT 100% USING IPV4 IN CALL FLOW IN MIXED IPV4/IPV6 TOPOLOGY Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 100% of the capacity of the MPEs. The topology will be using a mix of IPv4 and IPv6 addresses for OAM, signaling, and replication. Steps: DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 36 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 100% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-251: R-19526026-251 The general traffic model described in [R-19526026230] shall be repeated on hardware R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t Keywords: None 10.1.14 TEST CASE CAM-97266: MIXED CALL FLOW AT 100% USING IPV4 IN CALL FLOW WITH TOPOLOGY HIDING Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 100% of the capacity of the MPEs. For this test case both the Gx and Rx topology hiding features will be active. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Enable Gx and Rx topology hiding on the MRA DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 37 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 3. Prior to running the test, reset all counters on the MRA and MPEs 4. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 100% of the rated load of the MPEs 5. The traffic will run for 2 hours 6. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. The MRA and MPEs should process 100% of the calls generated by the traffic generator, with topology hiding for Gx and Rx active on the MRA. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-210: R-19526026-210 Reference Architecture B-1 shall support the following performance figure using BL 46 R-19526026-230: R-19526026-230 The mixed call flow model as defined in Section 2.3.14 will be run on hardware conf Keywords: Performance ANCHORAGE 10.1.15 TEST CASE CAM-94648: MIXED CALL FLOW AT 100% USING IPV4 IN CALL FLOW USING SCTP TRANSPORT Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 100% of the capacity of the MPEs. The transport used for this test will be SCTP. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 100% of the rated load of the MPEs DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 38 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-240: R-19526026-240 The general traffic model described in [R-19526026230] shall be repeated on hardware R-19526026-252: R-19526026-252 The general traffic model described in [R-19526026230] shall be repeated on hardware R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t Keywords: Performance ANCHORAGE 10.1.16 TEST CASE CAM-95432: BL460 MPE WITH 64GB RAM CAN SUPPORT 6.5 MILLION CONCURRENT SESSIONS IN B-1 TOPOLOGY Author: ppiacentini Summary: This test will verify that the HP BL460 C-Class MPE will be able to support up to 6.5 million active concurrent sessions and still provide the expected level of service. Steps: 1. Seagull will be used for the traffic generation. Traffic will be sent directly to an MPE cluster 2. Reboot the MPE cluster that will be used in this chassis to insure a clean setup to start the test 3. Two test drivers will be used for this test 4. The call flow used will be the call flow that was used for the performance tests with an SPR server for profile lookup 5. Between the 2 test drivers, run seagull scripts to generate Gx, Sd, and Rx sessions on the MPE, and calculate the hold time such that for a 1 hour period, there are a constant 6.5M active sessions, with the MPE's TPS at the DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 39 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY max rate of 6000 6. While the test is running, capture memory, CPU, and queue statistics from the MPE for future reference 7. While the test is running, monitor the KPI statistics screen for any unusual and unexpected issues on the MPE Expected Results: The HP BL460 C-Class MPE with 64GB RAM should be able to run normally with 6.5 million concurrent Gx, Sd, and Rx sessions. There should be no exceptions seen in the logs, or no abnormalities as seen in the KPI monitor screens Requirements R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t Keywords: Performance ANCHORAGE 10.1.17 TEST CASE CAM-95433: BL460 MRA WITH 64GB RAM CAN SUPPORT UP TO 52,000 TRANSACTIONS/SECOND IN B-1 TOPOLOGY Author: ppiacentini Summary: This test is to verify that the Oracle MRA running on an HP BL460 C-Class blade can support up to 52,000 transactions per second.when configured in the Reference Architechture B-1 topology. Test setup: (1) CMP cluster, (2) MRA clusters (2 sites), (10) MPE clusters (some simulated) 1. The MRA is configured for stateful routing, and should be evenly distributing the messages among the 10 MPES 2. The MPEs have Validation disabled, and there are no Data Sources enabled 3. The MPEs are configured to index by E164 address 4. All logging on all systems is set to WARN level Steps: 1. Reset all MRAs and MPES to insure a consistent start of test DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 40 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 2. Set up multiple seagull test drivers so that each instance of seagull uses a unique range of subscribers 3. Configure the seagull test drivers to send the following message flow: CCR-Initial, CCR-Update, CCRTerminate. 5. The traffic should run at steady-state rate of 52,000 for two hours 6. The logs should be monitored for any exceptions of unusual errors while the test is running 7. The KPI statistics screen will be monitored for any unexpected issues while the test is running 8. Latency statistics are captured for all of the calls Expected Results: The C-Class BL460 MRA should be able to support 52,000 tps total rate for a sustained period of time, while meeting the goal of 92% of the latency responses under 100ms and 99.9% of the latency responses under 200ms. Requirements R-19526026-210: R-19526026-210 Reference Architecture B-1 shall support the following performance figure using BL 46 Keywords: Performance ANCHORAGE 10.1.18 TEST CASE CAM-95619: BL460 MRA WITH 64GB RAM CAN SUPPORT UP TO 20M BINDINGS IN B-1 TOPOLOGY Author: ppiacentini Summary: This test is to verify that the Oracle MRA running on an HP BL460 C-Class blade can support up to 20M bindings in the Reference Architecture B-1 topology Test setup: (1) CMP cluster, (2) MRA clusters (2 sites), (10) MPE clusters 1. The MRA is configured for stateful routing, and should be evenly distributing the messages among the 10 MPES 2. The MPEs have Validation disabled, and there are no Data Sources enabled DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 41 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 3. The MPEs are configured to index by E164 address 4. All logging on all systems is set to WARN level Steps: 1. Reset all MRAs and MPES to insure a consistent start of test 2. Set up multiple seagull test drivers so that each instance of seagull uses a unique range of subscribers 3. Configure the seagull test drivers to send the following message flow: CCR-Initial, CCR-Update, CCRTerminate. 5. Generate enough sessions, using 20M unique Subscription-Ids, such that the MRA will contain 20M active bindings 6. Send a moderate amount of steady state traffic to the MRA containing the 20M bindings for a period of one hour 7. The logs should be monitored for any exceptions of unusual errors while the test is running 8. The KPI statistics screen will be monitored for any unexpected issues while the test is running. CPU and memory usage will be captured 9. Latency of the calls made during steady state period will be monitored Expected Results: The C-Class BL460 MRA with 64GB RAM should be able to support having 20M active bindings while handling steady state traffic with no system or forwarding problems seen. Requirements R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t Keywords: None 10.1.19 TEST CASE CAM-94646: LONGEVITY TEST - MIXED CALL FLOW AT 100% USING IPV4 IN CALL FLOW Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 100% of the capacity of the MPEs over an extended period of time. DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 42 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Session cleanup will be enabled for the duration of this test 4. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 100% of the rated load of the MPEs 5. The traffic will run for 72 hours without disruption 6. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. The MRA and MPE should process 100% of the calls generated by the traffic generator for the entire 72 hour period. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-230: R-19526026-230 The mixed call flow model as defined in Section 2.3.14 will be run on hardware conf R-19526026-252: R-19526026-252 The general traffic model described in [R-19526026230] shall be repeated on hardware R-19526026-260: R-19526026-260 The mixed call model defined in Section 2.3.14 shall be repeated on hardware configu R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t Keywords: Performance ANCHORAGE 10.1.20 TEST CASE CAM-95439: CMP CAN SUPPORT MANAGING UP TO 50 MPES/MRA GEOREDUNDANT CLUSTERS Author: ppiacentini Summary: This test case is to verify that the CMP can manage up to 50 MRAs and MPEs. All MRAs and MPEs are GeoRedundant clusters DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 43 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY Steps: 1. On the CMP, configure 50 MRAs and MPEs. This can be a combination of MRA and MPEs 2. Monitor the logs on the CMP for any unexpected errors while adding the MRAs and MPEs 3. In the CMP, go to the Policy Server/Configuration screen and verify that all the MPEs are showing up. If MRAs are configured verify that they show up in the MRA/Configuration screen 4. In the CMP, go into the statistics screens of several of the MRAs and MPES and verify that the statistics are updating 5. Go into the KPI Dashboard screen and verify that all MRAs and MPEs are represented, and the CMP shows statistics for each one Expected Results: The CMP can manage up to 50 MRA and MPE Geo-Redundant Clusters Requirements R-19526026-120: R-19526026-120 A S-CMP shall be able to manage and monitor a policy system of up to 50 MPE and MRA Keywords: Performance ANCHORAGE 10.1.21 TEST CASE CAM-95437: BL460 MRA WITH 64GB RAM CAN SUPPORT UP TO 8K CLIENT CONNECTIONS Author: ppiacentini Summary: This test is to verify that the Oracle MRA running on an HP C-Class BL460 blade with 64GB RAM can support up to 8k simultaneous and active Diameter client connections. Test setup: (1) CMP cluster, (1) MRA cluster (1) SDM cluster (10) MPE clusters 1. All connections from test drivers will terminate on the MRA 2. All traffic from the test drivers will be sent to the MRA over the Diameter connections 3. The MPEs are configured in Gx quota mode 4. The MPEs have Validation enabled, and has an SPR Data Source DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 44 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 5. The MPE is configured to index by E164 address 6. The test driver will use the same set of E164 addresses that are configured in the SPR 7. 10k unique Network Elements are defined in the CMP, and the MPEs are all associated with the 10k Network Elements Steps: 1. Set up the MRA so that it routes the Diameter traffic equally to the MPEs 2. Set up a seagull test driver so that it will generate sessions with the following call flow: Gx-Init, GxTerminate. Alternatively, diamcli's test rate could be used: 3. Set up the seagull scenario file so that it will bring up a total of 8k connections to the MRA. This will probably have to be spread out among several seagull test drivers 4. Set up the seagull script so that each simulated GGSN will send a small amount of traffic, say 10 CCRs/sec 5. Generate the traffic from all 10k clients for a period of 2 hours to the MRA/MPEs 6. Monitor the CPU, memory, and other statistics on the MRA and MPEs while the traffic is running 7. Check the KPI statistics and verify the connection count is correct, and the TPS to each MRA and MPE is accurate 8. Monitor the logs on the MRA and MPEs for any exceptions or other types of errors Sample test-rate command test rate gx -maxtransactions=100 -reportevery=1 -duration=3600 -delay=500 -transactionPool=10 numclients=1 -transactions=1 -peerCount=100 -identity=ggsn-1,ggsn-2,ggsn-3,ggsn-4,ggsn-5,ggsn-6,ggsn7,ggsn-8,ggsn-9,ggsn-10,ggsn-11,ggsn-12,ggsn-13,ggsn-14,ggsn-15,ggsn-16,ggsn-17,ggsn-18,ggsn-19,ggsn20,ggsn-21,ggsn-22,ggsn-23,ggsn-24,ggsn-25,ggsn-26,ggsn-27,ggsn-28,ggsn-29,ggsn-30,ggsn-31,ggsn-32,ggsn33,ggsn-34,ggsn-35,ggsn-36,ggsn-37,ggsn-38,ggsn-39,ggsn-40,ggsn-41,ggsn-42,ggsn-43,ggsn-44,ggsn-45,ggsn46,ggsn-47,ggsn-48,ggsn-49,ggsn-50,ggsn-51,ggsn-52,ggsn-53,ggsn-54,ggsn-55,ggsn-56,ggsn-57,ggsn-58,ggsn59,ggsn-60,ggsn-61,ggsn-62,ggsn-63,ggsn-64,ggsn-65,ggsn-66,ggsn-67,ggsn-68,ggsn-69,ggsn-70,ggsn-71,ggsn72,ggsn-73,ggsn-74,ggsn-75,ggsn-76,ggsn-77,ggsn-78,ggsn-79,ggsn-80,ggsn-81,ggsn-82,ggsn-83,ggsn-84,ggsn85,ggsn-86,ggsn-87,ggsn-88,ggsn-89,ggsn-90,ggsn-91,ggsn-92,ggsn-93,ggsn-94,ggsn-95,ggsn-96,ggsn-97,ggsn98,ggsn-99,ggsn-100 -remoteidentity=mpe26-59.camiant.com -realm=camiant.com -hostname=10.15.25.159 addressrange=1.0.0.0/18 -terminate=true -E164userids=3611000000/1000000 -subidtypes=E164 -ipcantype=0 calledstationid=mycalledstationid Expected Results: The BL460 C-Class MRA should be able to handle Diameter messages from up to 8k Diameter client connections simultaneously DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 45 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY Requirements R-19526026-510b: R-19526026-510b Up to 8000 Diameter Connections shall be set up to the MRA (stateful model) without Keywords: Performance ANCHORAGE 10.1.22 TEST CASE CAM-97374: MRA WILL SUPPORT 16 DIAMETER CONNECTIONS FROM A DRA Author: ppiacentini Summary: This test case is to verify that the MRA can support 16 connections from a DRA, with traffic being sent across all 16 connections Steps: 1. Configure a DRA to front-end an MRA containing 7 MPEs in its pool 2. Set up 16 connections between the DRA and the MRA 3. Run traffic at 100% load in this configuration for a period of two hours 4. The traffic will use the call model described in section 2.3.2 - LTE + Subscriber Quota (2 PDN connections), and the call model described in section 2.3.8 - OCS Throttling 5. Host-based routing will be used to route traffic bask to the AFs Expected Results: The MRA should be able to support traffic coming into it via 16 connections from a DRA. Requirements R-19526026-500b: R-19526026-500b Up to 16 Diameter Connections shall be set up between the DRA and MRA (stateful mode Keywords: Performance ANCHORAGE 10.1.23 TEST CASE CAM-98672: BASELINE STEP TEST - BL460-G8 MPE Author: ppiacentini Summary: This test is to get a baseline the basic code of the release, using a BL460-G8, with only Gx being used, and no DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 46 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY other Diameter interfaces. There is no validation, user lookup, policies being used. The only configuration on the MPE will be the Diameter Identity, realm, and Network Element used in the test. This should allow more direct apples-apples comparison between releases. The call flow will be simple Gx CCR-Initial, wait 10 secs, Gx CCR-Update, wait 10 secs, Gx CCR-Terminate Steps: 1. Configure a Diameter Identity and Realm on the MPE 2. Add the Network Elements that will be needed to run the test 3. Set up a seagull script that will send a Gx CCR-I, wait 10 secs, CCR-U, wait 10 secs, CCR-T 4. Send traffic to the MPE at a low level (ie, 200 calls/sec) which will generate 600 tps on the MRA. 5. Traffic will run for 15 minutes, script will capture protocol stats, CPU and Memory usage 6. Continually increase the traffic, running for 15 minutes at each level, until it reaches 200% of the rated load of the MPE The BL460-G8 MPE should handle all traffic up to 100% of its rated load without any problems There may be some load shedding seen as the traffic increases from 100% to 200% of the rated load. The latency should be within the guidelines specified in [R-19526026-130] for up to 100% load Requirements None Keywords: Performance ANCHORAGE 10.1.24 TEST CASE CAM-98673: BASELINE STEP TEST - BL460-G8 MRA Author: ppiacentini Summary: This test is to get a baseline the basic code of the release, using a BL460-G8, with only Gx being used, and no other Diameter interfaces. There is no validation, user lookup, policies being used. The only configuration on the MRA will be the Diameter Identity, realm, and Network Element used in the test. This should allow more direct apples-apples comparison between releases. The call flow will be simple Gx CCR-Initial, wait 10 secs, Gx CCR-Update, wait 10 secs, Gx CCR-Terminate Steps: DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 47 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 1. Configure a Diameter Identity and Realm on the MRA 2. Add the Network Elements that will be needed to run the test 3. The MRA will need at least 10 MPEs in its MPE pool, real or simulated 4. Set up a seagull script that will send a Gx CCR-I, wait 10 secs, CCR-U, wait 10 secs, CCR-T 5. Send traffic to the MPE at a low level (ie, 1500 calls/sec) which will generate 4500 tps on the MRA. 6. Traffic will run for 15 minutes, script will capture protocol stats, CPU and Memory usage 7. Continually increase the traffic and run at each level for 15 minutes until it reaches 200% of the rated load of the MPE The BL460-G8 MRA should handle all traffic up to 100% of its rated load without any problems There may be some load shedding seen as the traffic increases from 100% to 200% of the rated load. The latency should be within the guidelines specified in [R-19526026-130] for up to 100% load Requirements None Keywords: Performance ANCHORAGE 10.1.25 TEST CASE CAM-99004: BASELINE STEP TEST - BL460-G9 MPE Author: ppiacentini Summary: This test is to get a baseline of the basic code of the release, running on a BL460-G9, with only Gx being used, and no other Diameter interfaces. There is no validation, user lookup, policies being used. The only configuration on the MPE will be the Diameter Identity, realm, and Network Element used in the test. This should allow more direct apples-apples comparison between releases. The call flow will be simple Gx CCR-Initial, wait 10 secs, Gx CCR-Update, wait 10 secs, Gx CCR-Terminate Steps: 1. Configure a Diameter Identity and Realm on the MPE 2. Add the Network Elements that will be needed to run the test 3. Set up a seagull script that will send a Gx CCR-I, wait 10 secs, CCR-U, wait 10 secs, CCR-T DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 48 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 4. Send traffic to the MPE at a low level (ie, 200 calls/sec) which will generate 600 tps on the MRA. 5. Traffic will run for 15 minutes, script will capture protocol stats, CPU and Memory usage 6. Continually increase the traffic, running for 15 minutes at each level, until it reaches 200% of the rated load of the MPE The G9 MPE should handle all traffic up to 100% of its rated load without any problems There may be some load shedding seen as the traffic increases from 100% to 200% of the rated load. The latency should be within the guidelines specified in [R-19526026-130] for up to 100% load Requirements None Keywords: Performance ANCHORAGE 10.1.26 TEST CASE CAM-99005: BASELINE STEP TEST - BL460-G9 MRA Author: ppiacentini Summary: This test is to get a baseline the basic code of the release, using a BL460-G9, with only Gx being used, and no other Diameter interfaces. There is no validation, user lookup, policies being used. The only configuration on the MRA will be the Diameter Identity, realm, and Network Element used in the test. This should allow more direct apples-apples comparison between releases. The call flow will be simple Gx CCR-Initial, wait 10 secs, Gx CCR-Update, wait 10 secs, Gx CCR-Terminate Steps: 1. Configure a Diameter Identity and Realm on the MRA 2. Add the Network Elements that will be needed to run the test 3. The MRA will need at least 10 MPEs in its MPE pool, real or simulated 4. Set up a seagull script that will send a Gx CCR-I, wait 10 secs, CCR-U, wait 10 secs, CCR-T 5. Send traffic to the MPE at a low level (ie, 1500 calls/sec) which will generate 4500 tps on the MRA. 6. Traffic will run for 15 minutes, script will capture protocol stats, CPU and Memory usage 7. Continually increase the traffic and run at each level for 15 minutes until it reaches 200% of the rated load of DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 49 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY the MPE The BL460-G9 MRA should handle all traffic up to 100% of its rated load without any problems There may be some load shedding seen as the traffic increases from 100% to 200% of the rated load. The latency should be within the guidelines specified in [R-19526026-130] for up to 100% load Requirements None Keywords: Performance ANCHORAGE 10.1.27 TEST CASE CAM-98677: MIXED CALL FLOW AT 100% LOADS WITH OVER 1K POLICIES DEPLOYED Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-1 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 100% of the capacity of the MPEs, with the addition of at least 1000 additional policies beying deployed, above the number needed to support the mixed traffic flows. It is expected that the additional 1000 policies will be deployed, but not triggered. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. In addition to the policies being used for the mixed call flows, deploy at least 1000 additional policies, taken from the field if possible 4. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 100% of the rated load of the MPEs 5. The traffic will run for 2 hours 6. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 50 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY The latency should be within the guidelines specified in [R-19526026-130] Requirements None Keywords: Performance ANCHORAGE 10.1.28 TEST CASE CAM-98702: 2-SITE MRA AT 100% LOAD IN LEGACY MODE Author: ppiacentini Summary: This test is to verify that the 2 MRA sites in the Reference Architecture B-1 system can process loads equal to 100% of their capacity when configured as Primary and Backup in the Legacy mode Steps: 1. Two MRA sites are set up, with enough real or simulated MPEs in their pool to support a traffic load of 100% sent to the MRAs 2. The MRA association will be configured in "legacy" mode 3. Send traffic equal to 100% of the rated load to each of the two MRAs. A small amount of traffic should be routed between them 4. Traffic will run for a period of two hours 5. Monitor the CPU and Memory loads of the two MRAs during the test The two MRA sites should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements None Keywords: Performance ANCHORAGE 10.1.29 TEST CASE CAM-98703: 2-SITE MRA AT 100% LOAD IN ALGOV1 MODE Author: ppiacentini Summary: This test is to verify that the 2 MRA sites in the Reference Architecture B-1 system can process loads equal to DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 51 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 100% of their capacity when configured as Primary and Backup in the Algov1 mode Steps: 1. Two MRA sites are set up, with enough real or simulated MPEs in their pool to support a traffic load of 100% sent to the MRAs 2. The MRA association will be configured in "Algov1" mode 3. Send traffic equal to 100% of the rated load to each of the two MRAs. A small amount of traffic should be routed between them 4. Traffic will run for a perod of two hours 5. Monitor the CPU and Memory loads of the two MRAs during the test The two MRA sites should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements None Keywords: Performance ANCHORAGE 10.2 TEST SUITE : REFERENCE ARCHITECTURE B-2 Reference Architecture B-2 is an RMS based CMP, along with a full C-class chassis, consisting of one MRA cluster, and 7 MPE clusters. 10.2.1 TEST CASE CAM-95618: BL460 MRA WITH 64GB RAM CAN SUPPORT UP TO 52,000 TRANSACTIONS/SECOND IN B-2 TOPOLOGY Author: ppiacentini Summary: This test is to verify that the Oracle MRA running on an HP BL460 C-Class blade can support up to 52,000 transactions per second.when configured in the Reference Architechture B-2 topology. DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 52 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY Test setup: (1) CMP cluster, (1) MRA cluster1, (10) MPE clusters (some simulated) 1. The MRA is configured for stateful routing, and should be evenly distributing the messages among the 10 MPES 2. The MPEs have Validation disabled, and there are no Data Sources enabled 3. The MPEs are configured to index by E164 address 4. All logging on all systems is set to WARN level Steps: 1. Reset all MRAs and MPES to insure a consistent start of test 2. Set up mutliple seagull test drivers so that each instance of seagull uses a unique range of subscribers 3. Configure the seagull test drivers to send the following message flow: CCR-Initial, CCR-Update, CCRTerminate. 4. The traffic should run at steady-state rate of 52,000 for two hours 5. The logs should be monitored for any exceptions of unusual errors while the test is running 6. The KPI statistics screen will be monitored for any unexpected issues while the test is running 7. Latency statistics are captured for all of the calls Expected Results: The C-Class BL460 MRA should be able to support 52,000 tps total rate for a sustained period of time, while meeting the goal of 92% of the latency responses under 100ms and 99.9% of the latency responses under 200ms. Requirements R-19526026-300: R-19526026-300 Reference Architecture B-2 shall be supported with the following minimum physical har R-19526026-305: R-19526026-305 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-310: R-19526026-310 Reference Architecture B-2 shall support the following performance figure using BL 4 Keywords: Performance ANCHORAGE DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 53 of 94 PROPRIETARY INFORMATION 10.2.2 INTERNAL USE ONLY TEST CASE CAM-95435: BL460 MRA WITH 64GB RAM CAN SUPPORT UP TO 20M BINDINGS IN B-2 TOPOLOGY Author: ppiacentini Summary: This test is to verify that the Oracle MRA running on an HP BL460 C-Class blade can support up to 20M bindings in the Reference Architecture B-2 topology Test setup: (1) CMP cluster, (1) MRA cluster , (10) MPE clusters 1. The MRA is configured for stateful routing, and should be evenly distributing the messages among the 10 MPES 2. The MPEs have Validation disabled, and there are no Data Sources enabled 3. The MPEs are configured to index by E164 address 4. All logging on all systems is set to WARN level Steps: 1. Reset all MRAs and MPES to insure a consistent start of test 2. Set up mutliple seagull test drivers so that each instance of seagull uses a unique range of subscribers 3. Configure the seagull test drivers to send the following message flow: CCR-Initial, CCR-Update, CCRTerminate. 5. Generate enough sessions, using 20M unique Subscription-Ids, such that the MRA will contain 20M active bindings 6. Send a moderate amount of steady state traffic to the MRA containing the 20M bindings for a period of one hour 7. The logs should be monitored for any exceptions of unusual errors while the test is running 8. The KPI statistics screen will be monitored for any unexpected issues while the test is running. CPU and memory usage will be captured 9. Latency of the calls made during steady state period will be monitored Expected Results: The C-Class BL460 MRA with 64GB RAM should be able to support having 20M active bindings while handling steady state traffic with no system or forwarding problems seen. DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 54 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY Requirements R-19526026-300: R-19526026-300 Reference Architecture B-2 shall be supported with the following minimum physical har R-19526026-305: R-19526026-305 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-310: R-19526026-310 Reference Architecture B-2 shall support the following performance figure using BL 4 Keywords: Performance ANCHORAGE 10.3 TEST SUITE : REFERENCE ARCHITECTURE B-3 Reference Architecture B-3 consists of a CMP cluster, an MRA cluster, and 2 MPE clusters. All servers are Sun Netra X5-2 Rackmount servers 10.3.1 TEST CASE CAM-94649: MIXED CALL FLOW AT 80% USING IPV4 IN CALL FLOW Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-3 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 80% of the capacity of the MPEs. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 80% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. Expected Results: The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-400: R-19526026-400 Reference Architecture B-3 shall be supported with DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 55 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY the following minimum physical Co R-19526026-410: R-19526026-410 Reference Architecture B-3 shall support the following performance figures for Sun X5 Keywords: 10.3.2 Performance ANCHORAGE TEST CASE CAM-94650: MIXED CALL FLOW AT 100% USING IPV4 IN CALL FLOW Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-3 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 100% of the capacity of the MPEs. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 100% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-400: R-19526026-400 Reference Architecture B-3 shall be supported with the following minimum physical Co R-19526026-410: R-19526026-410 Reference Architecture B-3 shall support the following performance figures for Sun X5 Keywords: Performance ANCHORAGE 10.3.3 TEST CASE CAM-94651: MIXED CALL FLOW AT 120% USING IPV4 IN CALL FLOW Author: ppiacentini Summary: DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 56 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY This test is to verify that the Reference Architecture B-3 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 120% of the capacity of the MPEs. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 120% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. Expected Results: The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-400: R-19526026-400 Reference Architecture B-3 shall be supported with the following minimum physical Co R-19526026-410: R-19526026-410 Reference Architecture B-3 shall support the following performance figures for Sun X5 Keywords: Performance ANCHORAGE 10.3.4 TEST CASE CAM-94652: MIXED CALL FLOW AT 150% USING IPV4 IN CALL FLOW Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-3 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 150% of the capacity of the MPEs. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 150% of the rated load of the MPEs DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 57 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. 6. Verify that if load shedding kicks in, the correct levels and discards are used Expected Results: The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-400: R-19526026-400 Reference Architecture B-3 shall be supported with the following minimum physical Co R-19526026-410: R-19526026-410 Reference Architecture B-3 shall support the following performance figures for Sun X5 Keywords: Performance ANCHORAGE 10.3.5 TEST CASE CAM-94653: MIXED CALL FLOW AT 200% USING IPV4 IN CALL FLOW Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-3 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 200% of the capacity of the MPEs. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 200% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. 6. Verify that if load shedding kicks in, the correct levels and discards are used Expected Results: The MRA and MPE should process 100% of the calls generated by the traffic generator. DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 58 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-400: R-19526026-400 Reference Architecture B-3 shall be supported with the following minimum physical Co R-19526026-410: R-19526026-410 Reference Architecture B-3 shall support the following performance figures for Sun X5 Keywords: Performance ANCHORAGE 10.3.6 TEST CASE CAM-94654: LONGEVITY TEST - MIXED CALL FLOW AT 100% USING IPV4 IN CALL FLOW Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-3 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 100% of the capacity of the MPEs over an extended period of time. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 100% of the rated load of the MPEs 4. The traffic will run for 72 hours without disruption 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. The MRA and MPE should process 100% of the calls generated by the traffic generator for the entire 72 hour period. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-400: R-19526026-400 Reference Architecture B-3 shall be supported with the following minimum physical Co R-19526026-410: R-19526026-410 Reference Architecture B-3 shall support the following performance figures for Sun X5 Keywords: Performance ANCHORAGE DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 59 of 94 PROPRIETARY INFORMATION 10.3.7 INTERNAL USE ONLY TEST CASE CAM-94655: MIXED CALL FLOW AT 100% USING IPV4 IN CALL FLOW USING SCTP TRANSPORT Author: ppiacentini Summary: This test is to verify that the Reference Architecture B-3 system can process the requests generated by the mixed call flows as defined in Appendix-A at a load equal to 100% of the capacity of the MPEs. The transport used for this test will be SCTP. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the mixed call flows, so that the aggregate load generated by the full call cycle will equal 100% of the rated load of the MPEs 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-400: R-19526026-400 Reference Architecture B-3 shall be supported with the following minimum physical Co R-19526026-410: R-19526026-410 Reference Architecture B-3 shall support the following performance figures for Sun X5 Keywords: Performance ANCHORAGE 10.3.8 TEST CASE CAM-95434: ORACLE X5-2 MRA CAN SUPPORT UP TO 50,000 TRANSACTIONS/SECOND Author: ppiacentini Summary: DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 60 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY This test is to verify that the Oracle MRA running on an X5-2 can support up to 50,000 transactions per second. Test setup: (1) CMP cluster, (1) MRA cluster, (10) MPE clusters (some simulated) 1. The MRA is configured for stateful routing, and should be evenly distributing the messages among the 10 MPES 2. The MPEs have Validation disabled, and there are no Data Sources enabled 3. The MPEs are configured to index by E164 address 4. All logging on all systems is set to WARN level Steps: 1. Reset all MRAs and MPES to insure a consistent start of test 2. Set up multipleseagull test drivers so that each instance of seagull uses a unique range of subscribers 3. Configure the seagull test drivers to send the following message flow: CCR-Initial, CCR-Update, CCRTerminate. 5. The traffic should run at steady-state rate of 50,000 for an hour 6. The logs should be monitored for any exceptions of unusual errors while the test is running 7. The KPI statistics screen will be monitored for any unexpected issues while the test is running 8. At the conclusion of the test, the data_grinder.sh script will be used to trim off the ramp-up and ramp-down information, and calculate the latencies and percentiles using only the information from the steady-state period. The time to trim from the results would be 2 times the hold time (180 seconds) Expected Results: The Oracle X5-2 MRA should be able to support 50,000 tps total rate for a sustained period of time, while meeting the goal of 92% of the latency responses under 100ms and 99.9% of the latency responses under 200ms. Requirements R-19526026-400: R-19526026-400 Reference Architecture B-3 shall be supported with the following minimum physical Co R-19526026-410: R-19526026-410 Reference Architecture B-3 shall support the following performance figures for Sun X5 Keywords: Performance ANCHORAGE DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 61 of 94 PROPRIETARY INFORMATION 10.3.9 INTERNAL USE ONLY TEST CASE CAM-95431: ORACLE X5-2 MPE CAN SUPPORT 10 MILLION CONCURRENT SESSIONS Author: ppiacentini Summary: This test will verify that the Oracle X5-2 MPE will be able to support up to 10 million active concurrent sessions and still provide the expected level of service. The sessions will be a mix of Gx, Rx, and Sd sessions. Steps: 1. Seagull will be used for the traffic generation. Traffic will be sent directly to an MPE cluster 2. Reboot the MPE cluster that will be used in this chassis to insure a clean setup to start the test 3. Two test drivers will be used for this test 4. The call flow used will be the call flow that was used for the performance tests with an SPR server for profile lookup 5. Between the 2 test drivers, run seagull scripts to generate Gx, Sd, and Rx sessions on the MPE, and calculate the hold time such that for a 1 hour period, there are a constant 10M active sessions, with the MPE's TPS at the max rate of 6000 6. While the test is running, capture memory, CPU, and queue statistics from the MPE in a CSV file for future reference 7. While the test is running, monitor the KPI statistics screen for any unusual and unexpected issues on the MPE Expected Results: The Oracle X5-2 MPE should be able to run normally with 10 million concurrent Gx sessions. There should be no exceptions seen in the logs, or no abnormalities as seen in the KPI monitor screens Requirements R-19526026-400: R-19526026-400 Reference Architecture B-3 shall be supported with the following minimum physical Co R-19526026-410: R-19526026-410 Reference Architecture B-3 shall support the following performance figures for Sun X5 Keywords: Performance ANCHORAGE DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 62 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 10.3.10 TEST CASE CAM-95440: ORACLE X5-2 MRA CAN SUPPORT UP TO 40M BINDINGS Author: ppiacentini Summary: This test is to verify that the Oracle MRA running on a X352 server can support up to 40M bindings Test setup: (1) CMP cluster, (1) MRA cluster, (10) MPE clusters (some simulated) 1. The MRA is configured for stateful routing, and should be evenly distributing the messages among the 10 MPES 2. The MPEs have Validation disabled, and there are no Data Sources enabled 3. The MPEs are configured to index by E164 address 4. All logging on all systems is set to WARN level Steps: 1. Reset all MRAs and MPES to insure a consistent start of test 2. Set up mutliple seagull test drivers so that each instance of seagull uses a unique range of subscribers 3. Configure the seagull test drivers to send the following message flow: CCR-Initial, CCR-Update, CCRTerminate. 5. Generate enough sessions, using 20M unique Subscription-Ids, such that the MRA will contain 20M active bindings 6. Send a moderate amount of steady state traffic to the MRA containing the 20M bindings for a period of one hour 7. The logs should be monitored for any exceptions of unusual errors while the test is running 8. The KPI statistics screen will be monitored for any unexpected issues while the test is running. CPU and memory usage will be captured 9. Latency of the calls made during steady state period will be monitored Expected Results: The Oracle X5-2 MRA should be able to support having 40M active bindings while handling steady state traffic with no system or forwarding problems seen. Requirements R-19526026-400: R-19526026-400 Reference Architecture B-3 shall be supported with the following minimum physical Co R-19526026-410: R-19526026-410 Reference Architecture B-3 shall support the DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 63 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY following performance figures for Sun X5 Keywords: Performance ANCHORAGE 10.3.11 TEST CASE CAM-95438: ORACLE X5-2 MRA CAN SUPPORT UP TO 8K CLIENT CONNECTIONS Author: ppiacentini Summary: This test is to verify that the Oracle MRA running on a X5-2 can support up to 8k simultaneous and active Diameter client connections. Test setup: (1) CMP cluster, (1) MRA cluster (1) SDM cluster (10) MPE clusters (some simulated) 1. All connections from test drivers will terminate on the MRA 2. All traffic from the test drivers will be sent to the MRA over the Diameter connections 3. The MPEs are configured in Gx quota mode 4. The MPEs have Validation enabled, and has an SPR Data Source 5. The MPE is configured to index by E164 address 6. The test driver will use the same set of E164 addresses that are configured in the SPR 7. 1024 Unique Network Elements are defined in the CMP, and the MPEs are all associated with the 1024 unique Network Elements Steps: 1. Set up the MRA so that it routes the Diameter traffic equally to the MPEs 2. Set up a seagull test driver so that it will generate sessions with the following call flow: Gx-Init, GxTerminate. Alternatively, diamcli's test rate could be used: 3. Set up the seagull scenario file so that it will bring up a total of 8k connections to the MRA. This will probably have to be spread out among several seagull test drivers 4. Set up the seagull script so that each simulated GGSN will send a small amount of traffic, say 10 CCRs/sec 5. Generate the traffic from all 10k clients for a period of 2 hours to the MRA/MPEs 6. Monitor the CPU, memory, and other statistics on the MRA and MPEs while the traffic is running DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 64 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 7. Check the KPI statistics and verify the connection count is correct, and the TPS to each MRA and MPE is accurate 8. Monitor the logs on the MRA and MPEs for any exceptions or other types of errors Sample test-rate command test rate gx -maxtransactions=100 -reportevery=1 -duration=3600 -delay=500 -transactionPool=10 numclients=1 -transactions=1 -peerCount=100 -identity=ggsn-1,ggsn-2,ggsn-3,ggsn-4,ggsn-5,ggsn-6,ggsn7,ggsn-8,ggsn-9,ggsn-10,ggsn-11,ggsn-12,ggsn-13,ggsn-14,ggsn-15,ggsn-16,ggsn-17,ggsn-18,ggsn-19,ggsn20,ggsn-21,ggsn-22,ggsn-23,ggsn-24,ggsn-25,ggsn-26,ggsn-27,ggsn-28,ggsn-29,ggsn-30,ggsn-31,ggsn-32,ggsn33,ggsn-34,ggsn-35,ggsn-36,ggsn-37,ggsn-38,ggsn-39,ggsn-40,ggsn-41,ggsn-42,ggsn-43,ggsn-44,ggsn-45,ggsn46,ggsn-47,ggsn-48,ggsn-49,ggsn-50,ggsn-51,ggsn-52,ggsn-53,ggsn-54,ggsn-55,ggsn-56,ggsn-57,ggsn-58,ggsn59,ggsn-60,ggsn-61,ggsn-62,ggsn-63,ggsn-64,ggsn-65,ggsn-66,ggsn-67,ggsn-68,ggsn-69,ggsn-70,ggsn-71,ggsn72,ggsn-73,ggsn-74,ggsn-75,ggsn-76,ggsn-77,ggsn-78,ggsn-79,ggsn-80,ggsn-81,ggsn-82,ggsn-83,ggsn-84,ggsn85,ggsn-86,ggsn-87,ggsn-88,ggsn-89,ggsn-90,ggsn-91,ggsn-92,ggsn-93,ggsn-94,ggsn-95,ggsn-96,ggsn-97,ggsn98,ggsn-99,ggsn-100 -remoteidentity=mpe26-59.camiant.com -realm=camiant.com -hostname=10.15.25.159 addressrange=1.0.0.0/18 -terminate=true -E164userids=3611000000/1000000 -subidtypes=E164 -ipcantype=0 calledstationid=mycalledstationid Expected Results: The Oracle X5-2 MRA should be able to handle Diameter messages from up to 8k Diameter client connections simultaneously Requirements R-19526026-400: R-19526026-400 Reference Architecture B-3 shall be supported with the following minimum physical Co R-19526026-410: R-19526026-410 Reference Architecture B-3 shall support the following performance figures for Sun X5 Keywords: Performance ANCHORAGE 10.3.12 TEST CASE CAM-97382: VOLTE-PLUS-INBOUND-S9 ROAMING CALL FLOW AT 100% Author: ppiacentini Summary: Confirm that the Reference Architecture B-3 system can process the requests generated by the VoLTE-plusinbound-S9 call flows as defined in Appendix-A at a load equal to 100% of the capacity of the MPEs. Prerequisites: Reference Architecture B-3 Site configured as HPLMN: Global S9 configuration settings; Home and Visited Match Lists; Roaming Profile; Policy Server S9 Settings; DEA configured for MRA(s). SPR configured with IMSI-based Subscriber IDs DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 65 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY Another Site configured as VPLMN (originates S9 messages). Gx CCR-i messages must be contain both IMSI- and MSISDN-based subscriber IDs. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the VoLTE-plus-S9 call flows, so that the aggregate load generated by the full call cycle will equal 100% of the rated load of the MPEs. Note the prerequisite that the Gx CCR-i messages contain IMSI- and MSISDN-based subscriber IDs. 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. Expected Results: The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-252: R-19526026-252 The general traffic model described in [R-19526026230] shall be repeated on hardware R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t R-19526026-600: R-19526026-600 The following SPR/OCUDR Configurations and as shown in Table 15 shall be verified usi Keywords: Performance ANCHORAGE 10.3.13 TEST CASE CAM-97323: VOLTE-PLUS-OUTBOUND-S9 ROAMING CALL FLOW AT 100% Author: ppiacentini DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 66 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY Summary: Confirm that the Reference Architecture B-3 system can process the requests generated by the VoLTE-plusoutbound-S9 call flows as defined in Appendix-A at a load equal to 100% of the capacity of the MPEs. Prerequisites: Reference Architecture B-3 Site configured as VPLMN: Global S9 configuration settings; Home and Visited Match Lists; Roaming Profile; Policy Server S9 Settings; DEA configured for MRA(s). Another Site configured as HPLMN (recipient for S9 messages). Gx CCR-i messages must be contain both IMSI- and MSISDN-based subscriber IDs. Steps: 1. Preload all MPEs with 1M Gx sessions, using a range of E164 Subscription-Ids that will not be used by the scripts that are run in this test. Session cleanup is disabled 2. Prior to running the test, reset all counters on the MRA and MPEs 3. Start all scripts used for the VoLTE-plus-S9 call flows, so that the aggregate load generated by the full call cycle will equal 100% of the rated load of the MPEs. Note the prerequisite that the Gx CCR-i messages contain IMSI- and MSISDN-based subscriber IDs. 4. The traffic will run for 2 hours 5. Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. Expected Results: The MRA and MPE should process 100% of the calls generated by the traffic generator. During the time of the test there should be no exceptions seen on the servers, and no unexpected error or warning messages. The latency should be within the guidelines specified in [R-19526026-130] Requirements R-19526026-130: R-19526026-130 The policy systems described in Sections 2.4.3, 2.4.4, 2.4.5, and 2.4.6 while running R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad R-19526026-252: R-19526026-252 The general traffic model described in [R-19526026230] shall be repeated on hardware R-19526026-100: R-19526026-100 The Policy Component capacities shown in Table 2 shall be verified for each server t R-19526026-600: R-19526026-600 The following SPR/OCUDR Configurations and as shown in Table 15 shall be verified usi Keywords: Performance ANCHORAGE DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 67 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 10.4 TEST SUITE : REFERENCE ARCHITECTURE C-1 Reference Architecture C-1 will use the same Sun Netra X5-2 topology that is used in Reference Architecture B-3, except it will be configured for Wireline services and the major protocol being used will be RADIUS. For this set of tests, RADIUS sessions will be created on the Netra X5-2 MPEs behind a Netra X5-2 MRA. RADIUS sessions will be used for 90% of the calls, and the remaining 10% will be Gx sessions For 90% of the RADIUS sessions, the call flow will be Account-Start, wait for a period of 3 minutes, then Account-Stop. For 10% of the RADIUS calls, there will be a mid-session update triggered by a usage threshold on the Sd session. 10.4.1 CAM-94630:RADIUS WITH SD AND SPR PROFILE LOOKUP AT 80% LOAD Author: ppiacentini Summary: This test is to insure that the Netra X5-2 MPE will be able to handle a load of 100% of its rated capacity and stay within the latency and performance guidelines as specified in the Reference Architecture B document. The bulk of the traffic will be RADIUS sessions with quota management over Sd. A small amount of the traffic will be Gx sessions. . Steps: 1. Set up a Netra X5-2 MRA cluster, with at least one Netra X5-2 MPE cluster behind it 2. The MRA and MPE are configured for RADIUS, with a common shared secret 3. The MPE is configured with an SPR Data Source, and E164 will be used for the Subscription-Id 4. 10% of the total traffic will consist of Gx sessions 5. Policy will be in place to create an Sd session for each RADIUS session, and quota will be granted and tracked on the Sd session 6. Set up seagull test drivers such that they create and terminate sessions to generate an aggregate rate on the MPE that equals 80% of the rated load for this release. 80% of the RADIUS sessions will be Account-Start and Account-Stop, with a 3 minute DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 68 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY hold time. 10% of the sessions will have a mid-session update triggered by a usage threshold on the Sd session, and another 10% of the sessions will have a mid-sessions update due to a profile change on the SPR. Both of these mid-session updates will trigger a COA request to the BNG to install services. 7. Traffic will run at steady state for two hours 8. Protocol counts, CPU and memory usage will be monitored during the test Expected Results: The measured latency should not exceed the established guidelines described in the Reference Architecture B document. The MRA and MPE clusters should not have any failovers or exhibit any system or forwarding problems during the test RequirementsR-19526026-500: R-19526026-500 Reference Architecture C-1 shall be supported with the following minimum physical con R-19526026-510: R-19526026-510 ] Reference Architecture C-1 shall support the following performance figure using S R-19526026-530: R-19526026-530 The Policy Solution shall support the following general RADIUS and Sd message flows. R-19526026-540: R-19526026-540 The general traffic model for each MPE for the call flow for RADIUS COA and Sd per [ Keywords:Performance ANCHORAGE 10.4.2 CAM-94631:RADIUS WITH SD AND SPR PROFILE LOOKUP AT 100% LOAD Author: ppiacentini Summary: This test is to insure that the Netra X5-2 MPE will be able to handle a load of 100% of its rated capacity and stay within the latency and performance guidelines as specified in the Reference Architecture B document. The bulk of the traffic will be RADIUS sessions with quota management over Sd. A small amount of the traffic will be Gx sessions Steps: 1. Set up a Netra X5-2 MRA cluster, with at least one Netra X5-2 MPE cluster behind it 2. The MRA and MPE are configured for RADIUS, with a common shared secret 3. The MPE is configured with an SPR Data Source, and E164 will be used for the Subscription-Id 4. 10% of the total traffic will consist of Gx sessions 5. Policy will be in place to create an Sd session for each RADIUS session, and quota will be granted and tracked on the Sd session 6. Set up seagull test drivers such that they create and terminate sessions to generate an aggregate rate on the MPE that equals 100% of the rated load for this release. 80% of the RADIUS sessions will be Account-Start and Account-Stop, with a 3 minute hold time. 10% of the sessions will have a mid-session update triggered by a DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 69 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY usage threshold on the Sd session, and another 10% of the sessions will have a mid-sessions update due to a profile change on the SPR. Both of these mid-session updates will trigger a COA request to the BNG to install services. 7. Traffic will run at steady state for two hours 8. Protocol counts, CPU and memory usage will be monitored during the test Expected Results: The measured latency should not exceed the established guidelines described in the Reference Architecture B document. The MRA and MPE clusters should not have any failovers or exhibit any system or forwarding problems during the test Requirements R-19526026-500: R-19526026-500 Reference Architecture C-1 shall be supported with the following minimum physical con R-19526026-510: R-19526026-510 ] Reference Architecture C-1 shall support the following performance figure using S R-19526026-530: R-19526026-530 The Policy Solution shall support the following general RADIUS and Sd message flows. R-19526026-540: R-19526026-540 The general traffic model for each MPE for the call flow for RADIUS COA and Sd per [ Keywords: Performance ANCHORAGE 10.4.3 TEST CASE CAM-94632: RADIUS WITH SD AND SPR PROFILE LOOKUP AT 120% LOAD Author: ppiacentini Summary: This test is to insure that the Netra X5-2 MPE will be able to handle a load of 120% of its rated capacity and stay within the latency and performance guidelines as specified in the Reference Architecture B document. The bulk of the traffic will be RADIUS sessions with quota management over Sd. A small amount of the traffic will be Gx sessions. Steps: 1. Set up a Netra X5-2 MRA cluster, with at least one Netra X5-2 MPE cluster behind it 2. The MRA and MPE are configured for RADIUS, with a common shared secret 3. The MPE is configured with an SPR Data Source, and E164 will be used for the Subscription-Id 4. 10% of the total traffic will consist of Gx sessions DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 70 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 5. Policy will be in place to create an Sd session for each RADIUS session, and quota will be granted and tracked on the Sd session 6. Set up seagull test drivers such that they create and terminate sessions to generate an aggregate rate on the MPE that equals 120% of the rated load for this release. 80% of the RADIUS sessions will be Account-Start and Account-Stop, with a 3 minute hold time. 10% of the sessions will have a mid-session update triggered by a usage threshold on the Sd session, and another 10% of the sessions will have a mid-sessions update due to a profile change on the SPR. Both of these mid-session updates will trigger a COA request to the BNG to install services. 7. Traffic will run at steady state for two hours 8. Protocol counts, CPU and memory usage will be monitored during the test Expected Results: The measured latency should not exceed the established guidelines described in the Reference Architecture B document. The MRA and MPE clusters should not have any failovers or exhibit any system or forwarding problems during the test Requirements R-19526026-500: R-19526026-500 Reference Architecture C-1 shall be supported with the following minimum physical con R-19526026-510: R-19526026-510 ] Reference Architecture C-1 shall support the following performance figure using S R-19526026-530: R-19526026-530 The Policy Solution shall support the following general RADIUS and Sd message flows. R-19526026-540: R-19526026-540 The general traffic model for each MPE for the call flow for RADIUS COA and Sd per [ Keywords: Performance ANCHORAGE 10.4.4 TEST CASE CAM-94633: RADIUS WITH SD AND SPR PROFILE LOOKUP AT 150% LOAD Author: ppiacentini Summary: This test is to insure that the Netra X5-2 MPE will be able to handle a load of 150% of its rated capacity and stay within the latency and performance guidelines as specified in the Reference Architecture B document. The bulk of the traffic will be RADIUS sessions with quota management over Sd. A small amount of the traffic will be Gx sessions. Steps: 1. Set up a Netra X5-2 MRA cluster, with at least one Netra X5-2 MPE cluster behind it DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 71 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 2. The MRA and MPE are configured for RADIUS, with a common shared secret 3. The MPE is configured with an SPR Data Source, and E164 will be used for the Subscription-Id 4. 10% of the total traffic will consist of Gx sessions 5. Policy will be in place to create an Sd session for each RADIUS session, and quota will be granted and tracked on the Sd session 6. Set up seagull test drivers such that they create and terminate sessions to generate an aggregate rate on the MPE that equals 150% of the rated load for this release. 80% of the RADIUS sessions will be Account-Start and Account-Stop, with a 3 minute hold time. 10% of the sessions will have a mid-session update triggered by a usage threshold on the Sd session, and another 10% of the sessions will have a mid-sessions update due to a profile change on the SPR. Both of these mid-session updates will trigger a COA request to the BNG to install services. 7. Traffic will run at steady state for two hours 8. Protocol counts, CPU and memory usage will be monitored during the test Expected Results: The measured latency should not exceed the established guidelines described in the Reference Architecture B document. The MRA and MPE clusters should not have any failovers or exhibit any system or forwarding problems during the test Requirements R-19526026-500: R-19526026-500 Reference Architecture C-1 shall be supported with the following minimum physical con R-19526026-510: R-19526026-510 ] Reference Architecture C-1 shall support the following performance figure using S R-19526026-530: R-19526026-530 The Policy Solution shall support the following general RADIUS and Sd message flows. R-19526026-540: R-19526026-540 The general traffic model for each MPE for the call flow for RADIUS COA and Sd per [ Keywords: Performance ANCHORAGE 10.4.5 TEST CASE CAM-94634: RADIUS WITH SD AND SPR PROFILE LOOKUP AT 200% LOAD Author: ppiacentini Summary: This test is to insure that the Netra X5-2 MPE will be able to handle a load of 200% of its rated capacity and stay within the latency and performance guidelines as specified in the Reference Architecture B document. The bulk of the traffic will be RADIUS sessions with quota management over Sd. A small amount of the traffic will be Gx DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 72 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY sessions. Steps: 1. Set up a Netra X5-2 MRA cluster, with at least one Netra X5-2 MPE cluster behind it 2. The MRA and MPE are configured for RADIUS, with a common shared secret 3. The MPE is configured with an SPR Data Source, and E164 will be used for the Subscription-Id 4. 10% of the total traffic will consist of Gx sessions 5. Policy will be in place to create an Sd session for each RADIUS session, and quota will be granted and tracked on the Sd session 6. Set up seagull test drivers such that they create and terminate sessions to generate an aggregate rate on the MPE that equals 200% of the rated load for this release. 80% of the RADIUS sessions will be Account-Start and Account-Stop, with a 3 minute hold time. 10% of the sessions will have a mid-session update triggered by a usage threshold on the Sd session, and another 10% of the sessions will have a mid-sessions update due to a profile change on the SPR. Both of these mid-session updates will trigger a COA request to the BNG to install services. 7. Traffic will run at steady state for two hours 8. Protocol counts, CPU and memory usage will be monitored during the test Expected Results: The measured latency should not exceed the established guidelines described in the Reference Architecture B document. The MRA and MPE clusters should not have any failovers or exhibit any system or forwarding problems during the test Requirements R-19526026-500: R-19526026-500 Reference Architecture C-1 shall be supported with the following minimum physical con R-19526026-510: R-19526026-510 ] Reference Architecture C-1 shall support the following performance figure using S R-19526026-530: R-19526026-530 The Policy Solution shall support the following general RADIUS and Sd message flows. R-19526026-540: R-19526026-540 The general traffic model for each MPE for the call flow for RADIUS COA and Sd per [ Keywords: Performance ANCHORAGE 10.4.6 TEST CASE CAM-94635: RADIUS WITH SD AND SPR PROFILE LOOKUP LONGEVITY TEST DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 73 of 94 PROPRIETARY INFORMATION Author: INTERNAL USE ONLY ppiacentini Summary: This test is to insure that the Netra X5-2 MPE will be able to handle a load of 100% of its rated capacity for a period of at least 48 hours. Steps: 1. Set up a Netra X5-2 MRA cluster, with at least one Netra X5-2 MPE cluster behind it 2. The MRA and MPE are configured for RADIUS, with a common shared secret 3. The MPE is configured with an SPR Data Source, and E164 will be used for the Subscription-Id 4. Policy will be in place to create an Sd session for each RADIUS session, and quota will be granted and tracked on the Sd session 5, Session cleanup will be enabled for the duration of this test 6. Set up seagull test drivers such that they create and terminate sessions to generate an aggregate rate on the MPE that equals 80% of the rated load for this release. 80% of the sessions will be Account-Start and AccountStop, with a 3 minute hold time. 10% of the sessions will have a mid-session update triggered by a usage threshold on the Sd session, and another 10% of the sessions will have a mid-sessions update due to a profile change on the SPR. Both of these mid-session updates will trigger a COA request to the BNG to install services. 7. Traffic will run at steady state for 72 hours 8. Protocol counts, CPU and memory usage will be monitored during the test Expected Results: The Netra X5-2 will support RADIUS traffic for a period of at least 72 hours without any crashes, failovers, abnormal memory or CPU usage, or any other unexpected failures during the 72 hour period. Requirements R-19526026-530: R-19526026-530 The Policy Solution shall support the following general RADIUS and Sd message flows. R-19526026-540: R-19526026-540 The general traffic model for each MPE for the call flow for RADIUS COA and Sd per [ R-19526026-550: R-19526026-550 The Policy R12.1 system shall support 72 hour duration tests running Hardware Confi Keywords: Performance ANCHORAGE 10.5 TEST SUITE : REFERENCE ARCHITECTURE B IMPAIRMENT TESTS This test suite contains Fault Insertion test cases for Reference Architecture B 10.5.1 TEST CASE CAM-95400: BL460 C-CLASS ACTIVE MPE FAILOVER TO STANDBY MPE DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 74 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY AT 100% LOAD Author: ppiacentini Summary: This test is to verify that the BL460 C-Class MPE when faced with a 100% load can be failed over to the Standby MPE, which will proceed to process the load without issues Test setup: (1) CMP cluster, (1) SPR, (1) MRA cluster, (10) MPE clusters 1. The MPE is configured in Gx quota mode 2. The MPE has Validation enabled, and has an SPR Data Source 3. The MPE is configured to index by E164 address 4. The test driver will use the same set of E164 addresses that are configured in the SPR Steps: 1. Set up an MPE clusters so that the subscriber lookup will be to the SPR, and state and quote read/write will written to the SPR 2. Start traffic such that each MPE will be processing 100% of it's rated load at steady state 3. Fail over the active MPEs by doing a "service qp_procmgr restart" 4. Verify that the Standby MPE takes over as Active and can still process the traffic at 100% rate without any issues Expected Results: The Standby MPE should be able to become active and process 100% of the load in the case of a failure of the active MPE Requirements None Keywords: Performance Fault Insertion ANCHORAGE DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 75 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 10.5.2 TEST CASE CAM-99006: BL460-G9 C-CLASS ACTIVE MPE FAILOVER TO STANDBY MPE AT 100% LOAD Author: ppiacentini Summary: This test is to verify that the BL460-G9 C-Class MPE when faced with a 100% load can be failed over to the Standby MPE, which will proceed to process the load without issues Test setup: (1) CMP cluster, (1) SPR, (1) MRA cluster, (10) MPE clusters 1. The MPE is configured in Gx quota mode 2. The MPE has Validation enabled, and has an SPR Data Source 3. The MPE is configured to index by E164 address 4. The test driver will use the same set of E164 addresses that are configured in the SPR Steps: 1. Set up an MPE clusters so that the subscriber lookup will be to the SPR, and state and quote read/write will written to the SPR 2. Start traffic such that each MPE will be processing 100% of it's rated load at steady state 3. Fail over the active MPEs by doing a "service qp_procmgr restart" 4. Verify that the Standby MPE takes over as Active and can still process the traffic at 100% rate without any issues Expected Results: The Standby MPE should be able to become active and process 100% of the load in the case of a failure of the active MPE Requirements None Keywords: Performance Fault Insertion ANCHORAGE DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 76 of 94 PROPRIETARY INFORMATION 10.5.3 INTERNAL USE ONLY TEST CASE CAM-95401: ORACLE X5-2 ACTIVE MPE FAILOVER TO STANDBY MPE AT 100% LOAD Author: ppiacentini Summary: This test is to verify that the Oracle X5-2 MPE when faced with a 100% load can be failed over to the Standby MPE, which will proceed to process the load without issues Test setup: (1) CMP cluster, (1) SPR, (1) MRA cluster, (2) MPE clusters 1. The MPE is configured in Gx quota mode 2. The MPE has Validation enabled, and has an SPR Data Source 3. The MPE is configured to index by E164 address 4. The test driver will use the same set of E164 addresses that are configured in the SPR Steps: 1. Set up an MPE clusters so that the subscriber lookup will be to the SPR, and state and quote read/write will written to the SPR 2. Start traffic such that each MPE will be processing 100% of it's rated load at steady state 3. Fail over the active MPEs by doing a "service qp_procmgr restart" 4. Verify that the Standby MPE takes over as Active and can still process the traffic at 100% rate without any issues Expected Results: The Oracle X5-2 Standby MPE should be able to become active and process 100% of the load in the case of a failure of the active MPE Requirements None Keywords: Performance Fault Insertion ANCHORAGE 10.5.4 TEST CASE CAM-95402: BL460 C-CLASS GEO-REDUNDANT SITE1 MPE FAILOVER TO DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 77 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY SITE2 MPE AT 100% LOAD Author: ppiacentini Summary: This test is to verify that the BL460 C-Class Geo-Redundant MPE when faced with a 100% load can be failed over to the MPE, at the secondary site, which will proceed to process the load without issues Test setup: (1) CMP cluster, (1) SPR, (1) MRA cluster, (10) MPE clusters, including at least 1 Geo-Redundant cluster 1. The MPE is configured in Gx quota mode 2. The MPE has Validation enabled, and has an SPR Data Source 3. The MPE is configured to index by E164 address 4. The test driver will use the same set of E164 addresses that are configured in the SPR Steps: 1. Set up an MPE clusters so that the subscriber lookup will be to the SPR, and state and quote read/write will written to the SPR 2. Start traffic such that each MPE will be processing 100% of it's rated load at steady state 3. Fail over the active MPEs by doing a "service qp_procmgr restart" 4. Verify that the Standby MPE takes over as Active and can still process the traffic at 100% rate without any issues Expected Results: The secondary site MPE should be able to become active and process 100% of the load in the case of a failure of the MPE at the primary site Requirements None Keywords: Performance Fault Insertion ANCHORAGE 10.5.5 TEST CASE CAM-95403: ORACLE X5-2 GEO-REDUNDANT SITE1 MPE FAILOVER TO DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 78 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY SITE2 MPE AT 100% LOAD Author: ppiacentini Summary: This test is to verify that the Oracle X5-2 Geo-Redundant MPE when faced with a 100% load can be failed over to the MPE, at the secondary site, which will proceed to process the load without issues Test setup: (1) CMP cluster, (1) SPR, (1) MRA cluster, (2) MPE clusters, including at least 1 Netra GeoRedundant cluster 1. The MPE is configured in Gx quota mode 2. The MPE has Validation enabled, and has an SPR Data Source 3. The MPE is configured to index by E164 address 4. The test driver will use the same set of E164 addresses that are configured in the SPR Steps: 1. Set up an MPE clusters so that the subscriber lookup will be to the SPR, and state and quote read/write will written to the SPR 2. Start traffic such that each MPE will be processing 100% of it's rated load at steady state 3. Fail over the active MPEs by doing a "service qp_procmgr restart" 4. Verify that the Standby MPE takes over as Active and can still process the traffic at 100% rate without any issues Expected Results: The Oracle X5-2 secondary site MPE should be able to become active and process 100% of the load in the case of a failure of the MPE at the primary site Requirements None Keywords: Performance Fault Insertion ANCHORAGE 10.5.6 TEST CASE CAM-95404: BL460 C-CLASS ACTIVE MRA FAILOVER TO STANDBY MRA DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 79 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY AT 100% LOAD Author: ppiacentini Summary: This test is to verify that the BL460 C-Class MRA when faced with a 100% load can be failed over to the Standby MRA, which will proceed to process the load without issues Test setup: (1) CMP cluster, (1) SPR, (1) MRA cluster, (10) MPEs Steps: 1. Start traffic to the MRA so that it's running at 100% of the rated transaction rate for this release 2. Let the traffic run at steady state for at least 15 minutes 3. Fail over the active MRA by doing a "service qp_procmgr restart" 4. Verify that the Standby MRA takes over as Active and can still process the traffic at 100% rate without any issues Expected Results: The Standby MRA should be able to become active and process 100% of the load in the case of a failure of the active MRA Requirements None Keywords: Performance Fault Insertion ANCHORAGE 10.5.7 TEST CASE CAM-99007: BL460-G9 C-CLASS ACTIVE MRA FAILOVER TO STANDBY MRA AT 100% LOAD Author: ppiacentini Summary: This test is to verify that the BL460-G9 C-Class MRA when faced with a 100% load can be failed over to the Standby MRA, which will proceed to process the load without issues Test setup: (1) CMP cluster, (1) SPR, (1) MRA cluster, (10) MPEs Steps: DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 80 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 1. Start traffic to the MRA so that it's running at 100% of the rated transaction rate for this release 2. Let the traffic run at steady state for at least 15 minutes 3. Fail over the active MRA by doing a "service qp_procmgr restart" 4. Verify that the Standby MRA takes over as Active and can still process the traffic at 100% rate without any issues Expected Results: The Standby MRA should be able to become active and process 100% of the load in the case of a failure of the active MRA Requirements None Keywords: Performance Fault Insertion ANCHORAGE 10.5.8 TEST CASE CAM-95405: ORACLE X5-2 MRA FAILOVER TO STANDBY MRA AT 100% LOAD Author: ppiacentini Summary: This test is to verify that the Oracle X5-2 MRA when faced with a 100% load can be failed over to the Standby MRA, which will proceed to process the load without issues Test setup: (1) CMP cluster, (1) SPR, (1) Netra MRA cluster, (2) MPEs Steps: 1. Start traffic to the Oracle X5-2 MRA so that it's running at 100% of the rated transaction rate for this release 2. Let the traffic run at steady state for at least 15 minutes 3. Fail over the active MRA by doing a "service qp_procmgr restart" 4. Verify that the Standby MRA takes over as Active and can still process the traffic at 100% rate without any issues Expected Results: The Oracle X5-2 Standby MRA should be able to become active and process 100% of the load in the DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 81 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY case of a failure of the active MRA Requirements None Keywords: Performance Fault Insertion ANCHORAGE 10.5.9 TEST CASE CAM-95406: BL460 C-CLASS GEO-REDUNDANT PRIMARY SITE MRA FAILOVER TO SECONDARY SITE MRA AT 100% LOAD Author: ppiacentini Summary: This test is to verify that the Geo-Redundant BL460 C-Class MRA at the primary site when faced with a 100% load can be failed over to the secondary site MRA, which will proceed to process the load without issues Test setup: (1) CMP cluster, (1) SPR, (1) Geo-Redundant BL460 MRA cluster, (10) MPEs Steps: 1. Start traffic to the primary site BL460 MRA so that it's running at 100% of the rated transaction rate for this release 2. Let the traffic run at steady state for at least 15 minutes 3. Stop the Standby MRA at the primary site by doing "service qp_procmgr stop", and then fail over the primary site MRA by doing a "service qp_procmgr restart" 4. Verify that the secondary site MRA takes over as Active and can still process the traffic at 100% rate without any issues Expected Results: The secondary site BL460 MRA should be able to become active and process 100% of the load in the case of a failure of the primary site MRA cluster Requirements None Keywords: Performance Fault Insertion ANCHORAGE DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 82 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 10.5.10 TEST CASE CAM-95407: ORACLE X5-2 GEO-REDUNDANT PRIMARY SITE MRA FAILOVER TO SECONDARY SITE MRA AT 100% Author: ppiacentini Summary: This test is to verify that the Geo-Redundant Netra X5-2 MRA at the primary site when faced with a 100% load can be failed over to the secondary site MRA, which will proceed to process the load without issues Test setup: (1) CMP cluster, (1) SPR, (1) Geo-Redundant Oracle MRA cluster, (2) MPEs Steps: 1. Start traffic to the primary site Oracle MRA so that it's running at 100% of the rated transaction rate for this release 2. Let the traffic run at steady state for at least 15 minutes 3. Stop the Standby MRA at the primary site by doing "service qp_procmgr stop", and then fail over the primary site MRA by doing a "service qp_procmgr restart" 4. Verify that the secondary site MRA takes over as Active and can still process the traffic at 100% rate without any issues Expected Results: The secondary site Oracle X5-2 MRA should be able to become active and process 100% of the load in the case of a failure of the primary site MRA cluster Requirements None Keywords: Performance Fault Insertion ANCHORAGE 10.5.11 TEST CASE CAM-95408: SURGE OF SESSION CREATION REQUESTS Author: ppiacentini Summary: This test will verify that the MRA can withstand a burst of session creation attempts, such as the case where a gateway restarts and a surge of session requests is seen at the MRA DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 83 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY Steps: 1. Start seagull traffic to the MRA, such that the total tps is hitting the rated maximum for the MRA. Set the hold time of the sessions so that at steady state the MRA will be at or near its rated binding capacity. 2. Once the seagull is at a point where the above sessions are all terminating (should be at 50% of MRA tps capacity), start a new set of seaguls such that the tps of the session creations i(CCR_Inits) s now >70% of the rated tps capacity of the MRA. This should give a load on the MRA of 120%, where 50% is sessions terminating and 70% is sessions being created 3. On the new set of seagull scripts, try pause/unpause the traffic several times while the sessions are being created. 4. Try similar overloads with CCR-Updates, CCR-Terminates, and Rx messages Expected Results: The MRA should be able to handle surges of sessions being created that exceeds the number of sessions being terminated, where the total tps rate is up to 20% higher than the rated load for the MRA Requirements None Keywords: Performance Fault Insertion ANCHORAGE 10.5.12 TEST CASE CAM-98453: 100% LOAD TO MRA WITH 80MS LATENCY ADDED BETWEEN MRA AND MPE Author: ppiacentini Summary: This performance test will be using a configuration using the rest Of World configuration. In this configuration, an SPR will be used for Profile lookup and quota management In this test, 100% of the rated capacity for the chassis will be sent to the MRA, which will distribute the load evenly to the MPEs in it's pool. The chassis used will contain 7 Geo-Redundant MPE clusters, For this test an artificial latency of 80 ms will be introduced between the MRA and all of the MPEs. Test setup: (1) CMP cluster, (1) SPR, (1) Geo-Redundant MRA cluster, (7) Geo-Redudnat MPE clusters 1. The MPE is configured in Gx quota mode 2. The MPE has Validation enabled, and has an HSS Data Source 3. The MPE is configured to index by E164 address DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 84 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 4. The test driver will use the same set of E164 addresses that are configured in the SPR Steps: 1. Test drivers will be using seagull to generate the traffic to the MRA 2. The MRA will evenly distribute the traffic to the 7 MPEs in the chassis enclosure 3. Add 80ms latency between the MRA and the MPEs by using the following command example on the MRA, (changing the MPE IP addresses and bond interfaces appropriately): tc qdisc add dev bond0.240 root handle 1: prio tc qdisc add dev bond0.240 parent 1:3 handle 30: netem delay 30ms tc filter add dev bond0.240 protocol ip parent 1:0 prio 1 u32 match ip dst 10.15.25.159/32 flowid 1:3 4. Traffic will be sent to the MRA, so that it generates a 100% load on the MPEs in it's pool. Traffic will run for 2 hours. 5. During the test, the statistics of the MRA and the MPEs will be captured in a CSV file for memory usage, including CPU load, memory usage, protocol counters 6. During the test, the KPI statistics in the CMP will be monitored for CPU load and memory usage, and other information in real time 7 Remove the latency using the following commands on the MRA (again changing the bond interface and MPE IP): tc filter del dev bond0.240 protocol ip parent 1:0 prio 1 u32 match ip dst 10.15.25.159 flowid 1:3 tc qdisc del dev bond0.240 parent 1:3 handle 30: netem delay 30ms tc qdisc del dev bond0.240 root handle 1: prio Expected Results: 1. 99% of the call latencies will be below (100 + 80) ms 2. 99.9% of the call latencies will be below (200 + 80) ms 2. There will be no unexpected crashes, exceptions, or other unusual conditions on the MRA or MPEs Requirements None Keywords: Performance ANCHORAGE 10.5.13 TEST CASE CAM-98454: 100% LOAD TO MRA WITH 80MS LATENCY ADDED BETWEEN MPES AND SPR Author: ppiacentini Summary: This performance test will be using a configuration using the rest Of World configuration. In this configuration, an SPR DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 85 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY will be used for Profile lookup and quota management In this test, 100% of the rated capacity for the chassis will be sent to the MRA, which will distribute the load evenly to the MPEs in it's pool. The chassis used will contain 7 Geo-Redundant MPE clusters, For this test an artificial latency of 80 ms will be introduced between the MPEs and the SPR Test setup: (1) CMP cluster, (1) SPR, (1) Geo-Redundant MRA cluster, (7) Geo-Redudnat MPE clusters 1. The MPE is configured in Gx quota mode 2. The MPE has Validation enabled, and has an HSS Data Source 3. The MPE is configured to index by E164 address 4. The test driver will use the same set of E164 addresses that are configured in the SPR Steps: 1. Test drivers will be using seagull to generate the traffic to the MRA 2. The MRA will evenly distribute the traffic to the 7 MPEs in the chassis enclosure 3. Add 80ms latency between the MPEs and the SPR by using the following command example on the MPE, (changing the MPE IP addresses and bond interfaces appropriately): tc qdisc add dev bond0.240 root handle 1: prio tc qdisc add dev bond0.240 parent 1:3 handle 30: netem delay 30ms tc filter add dev bond0.240 protocol ip parent 1:0 prio 1 u32 match ip dst 10.15.25.159/32 flowid 1:3 4. Traffic will be sent to the MRA, so that it generates a 100% load on the MPEs in it's pool. Traffic will run for 2 hours. 5. During the test, the statistics of the MRA and the MPEs will be captured in a CSV file for memory usage, including CPU load, memory usage, protocol counters 6. During the test, the KPI statistics in the CMP will be monitored for CPU load and memory usage, and other information in real time 7 Remove the latency using the following commands on the MRA (again changing the bond interface and MPE IP): tc filter del dev bond0.240 protocol ip parent 1:0 prio 1 u32 match ip dst 10.15.25.159 flowid 1:3 tc qdisc del dev bond0.240 parent 1:3 handle 30: netem delay 30ms tc qdisc del dev bond0.240 root handle 1: prio Expected Results: 1. 99% of the call latencies will be below (100 + 80) ms 2. 99.9% of the call latencies will be below (200 + 80) ms 2. There will be no unexpected crashes, exceptions, or other unusual conditions on the MRA or MPEs DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 86 of 94 PROPRIETARY INFORMATION Requirements None Keywords: Performance ANCHORAGE INTERNAL USE ONLY 10.5.14 TEST CASE CAM-99096: 100% LOAD TO 2-SITE MRAS WITH DROPPED DRMA MESSAGES BETWEEN MRAS Author: ppiacentini Summary: This performance test will be using a configuration using the rest Of World configuration. In this configuration, there will be 2 MRA sites. Each site will be receiving 100% traffic load, A small number DRMA messages between the 2 MRA sites will be dropped. For this test 1/10th of a percent of the DRMA messages between the MRAs will be dropped Test setup: (1) CMP cluster, (2) MRA sites (Primary/Backup), sufficient MPEs or simulators behind each MRA Steps: 1. Test drivers will be using seagull to generate the traffic to the 2 MRAs 2. Set up the MRAs so that .1% of the DRMA messages will be dropped If using the "tc" command: tc qdisc add dev bond1.240 root handle 1: prio tc qdisc add dev bond1.240 parent 1:3 handle 30: netem loss 0.1% of 1000) packets to be randomly dropped) (This causes 1/10th of a percent (i.e. 1 out tc filter add dev bond1.240 protocol ip parent 1:0 prio 1 u32 match ip dst 10.15.25.197 flowid 1:3 (to undo): tc qdisc del dev bond1.240 root handle 1: prio Alternatively, iptables may be used to do this: iptables -A INPUT --dport 3868 -m statistic --mode random --probability 0.01 -j DROP DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 87 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY 3. Using seagull, send traffic to both MRAs so that the load is 100% of the rated capacity of the MRA 4. Monitor the success of the calls to both MRAs 5. Verify that the DBR message timeout counters are increasing, but has no detremental effect on the MRAs Expected Results: The MRAs will not have any noticeable problems when a small number of DRB messeges are lost between them Requirements None Keywords: Performance ANCHORAGE DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 88 of 94 PROPRIETARY INFORMATION APPENDIX A. Review Type: Document: Date: Author: Moderator: INTERNAL USE ONLY REVIEW SUMMARIES On-Line TP010213 v0.1 05/01/2015 Paul Piacentini NA Department Reviewer Reviewer Type S Matthew Buehler CUSTOMER NPX Gregg Satkamp CUSTOMER CS Vito Politano SME D Joe Bearden SME TPM Christopher Berluti SME Q Mark Young SME PM Ben Crane SME SA Tarek Abou Assali SME S- Sustaining Piriyakala Suresh SME Vote Comments: jbearden: Please mark Cust Doc as N/A for this review. ppiacent: removed cost doc from the review pisukapa: I guess, including the sample startup scripts for the testcase scenarios and policies for various call flows would help in excution ppiacent: Most of the script work and policy development will be done after this test plan is completed. I can include a few of the policies that I have so far in the test plan for examples. ppiacent: Added a sample of the policies that will be used in Call #1 pisukapa: Teh callflow for "LTE+ Subscriber Quota(2 PDN Connections) Separate Quota Reporting and Revalidation Timeout" , is missing? or ..? pisukapa: Also, for "VoLTE +S9 Outbound Roaming" AND "VoLTE + S9 Inbound Roaming" are missing? ppiacent: The "LTE+ Subscriber Quota(2 PDN Connections) Separate Quota Reporting and Revalidation Timeout" is Call2. Allot GW sends Revalidation Timeout first, then quota update. The S9 Inbound and Outbound are going to use the same VoLTE script as Call6, if possible. I will add a mention of that. DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 89 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY pisukapa: Are there any (sample)policies used/deployed for the call flow tests? to satisfy the gnereal goals of the tests ? ppiacent: The work on script development and policy definitions will be done after this test plan is written. I can include a few policies that were written so far for examples. ppiacent: added a sample of policies to be used for Call #1 pisukapa: Apart from the listed Latency Goals listed As per the FE0077538, the Latecy Requirements are "3. The transaction response time of the Policy system with two sites for 92% of all transactions shall not exceed 167 + (2 * LL) ms; and 4. The transaction response time of the Policy system with two sites for 99.9% of all transactions shall not exceed 250 + (2 * LL) ms, where LL is the latency of the link between the two systems 5. The latency from the SPR is expected and assumed to be 25ms or less as part of the transaction. 6. The latency from the OCS simulator is expected and assumed to be 20ms or less as part of the transaction 7. The latency from the PCEF and AF simulator is expected and assumed to be 20ms or less as part of the transaction. " so wondering do u need the above golas as well? ppiacent: added these to section 9.2 pisukapa: "Start all scripts used for the mixed call flows" - can include/publish the start scripts in the test plan? ppiacent: Most of the scripts that will be needed have not been written yet. Work on creating those will be done after this test plan is completed. ppiacent: Added a sample of the policies that will be used for Call #1 pisukapa: I guess, this comment belongs for all the TestCases: Monitor the CPU, memory, call latency, other KPIs will be collected and logs of all servers for the duration of the test. "ther KPIs will be collected" is missing in step # 5 ppiacent: Changed the related step in all applicable test cases to" "Monitor the CPU, memory, and logs of all servers for the duration of the test. CPU, Memory, call latencies, average session size, and other KPIs will be recorded during the test. " isukapa: The Requirements : section, the sentences are CUT and It seems, Incomplete as well. For example: R-19526026-200: R-19526026-200 Reference Architecture B-1 shall be supported with the following minimum physical har R-19526026-205: R-19526026-205 The Policy DL460 G8 blade configuration used for testing MPE and MRA shall include ad ppiacent: The test case repository tool (TestLink) truncates the requirement due to a field length limitation. The full requirement description can be looked up either by logging into TestLink and viewing the full description of the requirement, or by looking up the requirement in FE007538 DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 90 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY pisukapa: TYPO: "desribed" pisukapa: described ppiacent: corrected spelling pisukapa: The Requirements mappings are missing ppiacent: There are no requirements in FE007538 related to impairment tests and recovery. These are more like general expectations. pravrajp: TYPO: "6.500" pravrajp: Should this be 6500? ppiacent: corrected pravrajp: Also apply to whole document wherever space is required to be inserted ppiacent: Checked and corrected formatting in steps for all test cases pravrajp: Please insert space also apply to whole document wherever space is to be inserted ppiacent: Checked and corrected formatting in steps for all test cases pravrajp: This para is the expected result so please put this under Expected Result Row ppiacent: Checked and corrected formatting in steps for all test cases pravrajp: Please put this paragraph in the separate Expected Result row ppiacent: Corrected paragraph header pravrajp: Same as my previous two comments ppiacent: Checked and corrected formatting in steps for all test cases pravrajp: Same as pervious comment ppiacent: Checked and corrected formatting in steps for all test cases pravrajp: Same as my pervious comment ppiacent: Checked and corrected formatting in steps for all test cases pravrajp: Same as my pervious comment; also apply to whole document wherever required. ppiacent: Checked and corrected formatting in steps for all test cases pravrajp: TYPO: "wil" ppiacent: Corrected spelling pravrajp: TYPO: "wil" ppiacent: Corrected spelling DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 91 of 94 PROPRIETARY INFORMATION Review Type: Document: Date: Author: Moderator: INTERNAL USE ONLY In-Person TP010213 v0.2 05/13/2015 Paul Piacentini NA Department Reviewer Reviewer Type S Matthew Buehler CUSTOMER NPX Gregg Satkamp CUSTOMER CS Vito Politano SME D Joe Bearden SME TPM Christopher Berluti SME Q Mark Young SME PM Ben Crane SME SA Tarek Abou Assali SME S- Sustaining Piriyakala Suresh SME Vote 3 (Prasad) N/A Comments: Paul: Do we need to add the tests using “tc” command for latency back in Yes, they will be added to test plan Piriya: need a test with large number of policies Will create one test case at 100% load using large number of policies Jared/Mat: need to have 2 MRA sites, not just one Topology B-1 will have 2 MRA sites Sayan: add host based routing Will add host based routing to 16 connection test Sayan: use IMSI for LTE tests Will add IMSI to LTE calls Mat: are the mezzanine cards needed? DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 92 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY Chris B: yes a customer is using them (will add them to all blades) Aruna: is a large SPR profile needed? Will document average profile size if possible Piriya: run session cleanup during longevity test Longevity test will have session cleanup enabled with the default parameters Sayan: need to run just the 100% S9 inbound/outbound tests Removed all S9 tests except the 100% inbound and outbound Piriya: Subscriber Activity log on for one user during tests One subscriber will be added to subscriber trace for all tests Piriya: document average session size during tests The average session size will be documented Paul: add baseline step tests for the MRA and MPE Added baseline step tests for MRA and MPE Sayan: add Gx sessions to RADIUS tests Added some Gx traffic to the RADIUS tests Review Type: Document: Date: Author: Moderator: In-Person TP010213 v0.3 06/03/2015 Paul Piacentini NA Department Reviewer Reviewer Type S Matthew Buehler CUSTOMER NPX Gregg Satkamp CUSTOMER CS Vito Politano SME D Joe Bearden SME TPM Christopher Berluti SME Q Mark Young SME PM Ben Crane SME SA Tarek Abou Assali SME S- Sustaining Piriyakala Suresh SME DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Vote N/A Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 93 of 94 PROPRIETARY INFORMATION INTERNAL USE ONLY Jared: add test case where some amount pf messages between MRAs are dropped Added test case to drop .1% of messages between MRAs at 100% load Bash: add some cases for BL460-G9 blades Added 4 test cases for BL460-G9 Review Type: Document: Date: Author: Moderator: In-Person TP010213 v0.5 06/17/2015 Paul Piacentini NA Department Reviewer Reviewer Type Vote S Matthew Buehler CUSTOMER 3 NPX Gregg Satkamp CUSTOMER 3 CS Vito Politano SME D Joe Bearden SME TPM Christopher Berluti SME Q Mark Young SME PM Ben Crane SME SA Tarek Abou Assali SME S- Sustaining Piriyakala Suresh SME N/A Received votes of 3 from Guarav Paliwal and Prasad Isukapalli DO NOT COPY. Hard copies of this document are for REFERENCE ONLY. Get the latest version from Documentum. Title: Reference Architecture B Performance test plan Filename: TP10213.docx REV: 1.0 Page 94 of 94