CONSULT21 BRIEFING 21CN PSTN emulation technical issues briefing Issue1: Briefing Number C21–NS-002 Date: 17th February 2006 IMPORTANT NOTE Please note that this represents current thinking and is subject to change. This document is supplied to you in confidence as part of the Consult21 Process. It should not be circulated outside the agreed Consult21 Communications Providers. The information contained within this document is provided by BT to promote understanding and facilitate ongoing discussions at the Network Structure Working Group. Any questions relating to this information should be raised initially with this forum or by emailing consult.21@bt.com Copyright © British Telecommunications plc 2006 Page 1 Background This document summarises the BT view on the four issues identified in the PSTN emulation technical consultation document and referred to the Consult21 Network Structures Working Group. 1 Managing voice bandwidth on the MSIL The bandwidth on the MSIL between Border Gateways or equivalent in each CP’s network needs to be assigned to the different services. How should this assignment be done and policed? This issue will be referred to the Consult21 Network Structures Working Group for resolution. No response required for this consultation. BT view The MSIL link, as defined in the NICC standards, can offer different transport capabilities to match different service types e.g. TDM supporting services such as traditional PSTN circuit-switched interconnects and PPCs ATM supporting services such as DataStream Ethernet -yet to be defined but with a view of supporting Ethernet services in future IP - for a number of services such as Internet Peering, PSTNoIP, VPNs etc. Each service that is offered by BT has to define how it uses the appropriate MSIL transport capability; the transport capability of an MSIL is useless by itself. The NICC standard requires that the each Ethernet VLAN is of a defined fixed bandwidth which has policing and traffic shaping at each end. In 21CN this is a requirement on the Fibre MSAN Ethernet switch as well as the management control to set it up. For PSTNoIP there are two service types, signalling and media: Signalling - PSTN SIP-I signalling will have its own fixed bandwidth VLAN in a MSIL. Any CP can have multiple physical PoHs with one (and possibly, but unlikely, more) signalling VLANs in each one. SIP-I Signalling over SCTP between two Call Servers can be sent over more than one VLAN on different geographic POHs with automatic failure and recovery to get geographic resilience. A signalling VLAN is capable and likely to simultaneously carry the signalling traffic between a number of different CS<->CS pairs. See embedded diagram below from the NICC standard. The bandwidth required is therefore dependent on the number of CSs sharing the bandwidth of each VALN, , the geographic fail-over strategy for each CS to CS signalling path they choose and the peak signalling rates (+ some overhead) and is impossible to specify as a rule of thumb. Maybe the NICC management working group should produce some planning guidlines for this and the media stream. The key thing to note is that the signalling service and its physical implementation is independent to the media steam for PSTNoIP or any other service. Media - The PSTNoIP media stream is carried on its own VLAN within the MSIL of fixed policed bandwidth. Each media stream (telephone call) passes through a border gateway controlled by its own call server / SIP back-to-back user agent. As a result of the signalling requests, the border gateway is instructed to open up a pinhole for that individual media stream and the gateway polices the bandwidth that is requested. (See embedded diagram below). Each media stream will require approximately 100kbit/s for each 64kbit/s speech stream due to the packet overheads. There should also be some overhead left in the fill of the fixed VLAN bandwidth as a guide. This overhead depends on the size of the VLAN, small ones needing greater percentage overhead than larger ones. Page 2 Note that there is a bandwidth management function in this diagram which relates only to the interconnect. This function is can be provided by the media border function which can refuse requests from the " Edge session controller" (i.e. call server / SIP back-to-back user agent) to open a policed media stream pin hole if there is insufficient VLAN Bandwidth. 2 Non functional aspects The capacity of traffic passing between networks in the new technology appears to be flexible in that, up to a certain point, the interconnect can handle an amount of calls, yet to be determined. At some point, the number of calls per minute will reach a maximum and overload arrangements will need to be employed. Also during the planning phase, interconnect resilience should be addressed. These issues will be referred to the Consult21 Network Structures Working Group for resolution. No response required for this consultation. BT view The ability to prevent more calls than bandwidth (i.e. "route busy"), the overload of call processing (i.e. rate of calls per minute), and resilience scenarios are addressed in the NICC doc ND1612 and SPEC 17. Links between networks must be bandwidth managed (as discussed under Managing voice bandwidth on the MISL). SIP-I must support ACC in the ISUP for overload control. The annexes of ND1612 show signalling resilience options but no figures as to how reliable an Interconnect should be. Some (contractual) discussions will be needed on the agreed resilience and how this is achieved 3 Impact on Call Servers Call Servers have a finite call handling capacity which needs to be planned and policed. The impact of call volumes and call processing rates on the servers needs to be agreed for the purpose of interconnect. This issue will be referred to the Consult21 Network Structures Working Group for resolution. No response required for this consultation. BT view In general, the calls capacity of the network is driven by two parameters: 1. The amount of bandwidth required to carry the media (64kbit/s per call in TDM and around 100kbit/s per call in IP) 2. The number of calls per second that the Call Servers can process. There is a relationship between the two that is determined by the call holding time. For example, if the call holding time is 120 seconds, then each circuit supports 1 call every 120s (+ a bit more as circuits can't be released and seized instantly), hence 120 circuits would produce 1 call per second, 1200 circuits would produce 10 calls per second, etc. This relationship can be broken by traffic patterns that have very short call hold times, e.g. televotes. If the call holding time is reduced then a given number of circuits produce more calls per second. Page 3 In IP there are no circuits but we know that a call uses a given amount of bandwidth hence a similar relationship exists between bandwidth, call holding time and calls per second. Such dimensioning calculations also have to take account of short calls (e.g. Televotes). We also have to ensure that the signalling bandwidth between components is sufficient to cope with the required calls per second. It will get more complicated when we move to variable bandwidth calls - but for PSTN/ISDN the bandwidth is fixed. For PSTN/ISDN, the Call Server will use bandwidth management, initially in the form of call counting, to ensure sufficient bandwidth between media components and the core IP network before it sets up a call. The 21CN core network is being deployed to be non-blocking i.e. more than sufficient bandwidth available. Some Session Border Controllers and Call Servers have a call rate limiter function both inbound and outbound. This will help but an overload control mechanism such as ACC will be needed. SIP has nothing yet. BT working in TISPAN for an NGN overload control protocol (GOCAP) which could be standardised late summer. 4 Circuit reservation for 999/112 calls In the current legacy network circuits are reserved for the use of 999/112 emergency services. CPs who are interested in using the NGN equivalent to this reservation of capacity should note that a similar call counting or circuit reservation will need to be incorporated in their NGN design and BT interconnect for 999/112 calls. For information only. No response required for this consultation BT view See the embedded document that is BT’s proposed text on Priority Calls to be included in ND1612. Note that it is planned to expose this text to the wider NICC/TSG membership during w/c 20 February. D:\Documents and Settings\802045124\My Documents\Voice Technology\Standards\NICC\NDs for IP IX\ND1612 PSTN-ISDN Gen Connectivity\Comments\ND1612 Pr Page 4