IBM Software Group
IBM CICS® Transaction Server for z/OS™ V3.1
Performance Report
April 2005
CICS Performance Group
IBM Hursley
© 2004 IBM Corporation
IBM Software Group
Performance summary (V3.1 compared to V2.3) Highlights

Traditional VSAM (DSW workload)
–

Traditional DB2 (RTW workload)
–

Shorter pathlength for C++ method linkage
Any benefits will depend on size, number of calls, depth of nested calls etc.
OPENAPI
–
–
2
Program call using CONTAINER, within 2% CPU/Tran of COMMAREA linkage on 1 MIP application
XPLINK
–
–

Using WSDL2CICS, our workload showed up to 20% CPU saving compared to CTS 2.3 SOAP feature
Containers
–

Concurrent session limit raised from 250 to 65000 per region
SOAP
–

Concurrent session limit raised from 900 to 65000 per region
Some customers will start using persistent sessions and then also make CPU savings
HTTPS
–

CPU used - Equivalent
HTTP
–
–

CPU used - Equivalent
Java
–

CPU used - Equivalent
Applications other than those that call DB2 can now exploit OPEN TCBs
Throughput not limited by QR TCB, multiple CPUs can now be exploited concurrently.
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Hursley Performance Environment
 Exclusive use LPAR on Z990 2084 332
– 3 dedicated CPUs for system under test
 Shared use of LPAR for network simulation
 Use LSPR to scale CPU figures to other target systems
– 2084 332 ITR/32
comparison?
or
2084 303 ITR/3 for CP speed
• We found the 303/3 to be more accurate for scaling workloads
3
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Release/Release
4
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Release/Release Comparisons
 COBOL VSAM
– 'DSW ' workload
 COBOL DB2
– 'RTW' workload
 Java
– Standard Java/JCICS and EJB workload
 Compare ITR (tran per CPU second)
– Transaction rate in TORs/Total CICS regions CPU
 Taken from average of 5 Transaction rates
5
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Release to Release - DSW Workload (VSAM)
 Hardware
– 2084-303 system under test
– 2084-303 TPNS driver
 Software
– z/OS 1.6
 CICS Environment
 Workload
– 2 TORs, 2 AORs, 1 FOR all MRO connected
– 34 transaction types
– All VSAM files recoverable
– COBOL applications
– VTAM HPO used
– 40% transactions invoke menu
– No transaction isolation or storage protection
– 32 VSAM KSDS files
– Long running mirror in FOR
– average 6 FC calls per transaction
– EXEC CICS LINK to program for each FC call
– 69% Read, 10% Read for Update, 9% Update,
– 11% Add , 1% Delete
– XEIIN and XEIOUT enabled
– TRUE enabled
– CICS Performance Monitoring on
6
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
DSW Performance Data
ITR
ETR
Average ITR = 2809
384.99
531.50
708.11
1055.21
1586.01
TORs%
7.9
10.6
14.2
20.9
31.6
AORs%
23.6
32.2
42.4
62.7
92.8
FOR%
10.9
14.8
19.0
28.5
41.5
Resp
0.037
0.036
0.039
0.036
0.040
ms/tran
1.10
1.08
1.06
1.06
1.04
2727
2777
V2.3
2830
2830
2884
ITR
ETR
Average ITR = 2794
Delta = 0.5%
7
384.93
531.49
710.50
1053.94
1564.69
TORs%
8.0
10.8
14.2
20.8
30.8
IBM CICS TS for z/OS V3.1 | Performance Report
AORs%
23.9
32.5
43.1
63.2
92.0
FOR%
11.0
14.9
19.3
28.5
40.7
Resp
0.037
0.036
0.035
0.037
0.043
ms/tran
1.11
1.09
1.07
1.06
1.04
2702
2752
V3.1
2803
2830
2884
© 2003 IBM Corporation
IBM Software Group
Release to Release - RTW Workload (DB2)
 Hardware
– 2084- 303 system under test
 Software
– z/OS 1.6
– 2084- 303 TPNS driver
 Workload
 CICS Environment
– Single Region
– 7 transaction types
– COBOL applications - Not Thread safe
– 20 Database tables
 DB2 Environment
– 7.1.0
– Thread priority = High
– Average 200 DB2 calls per transaction
– 54% Select, 1% inset, 1% update,
1%delete,
– Protect num = 0
– 8% open cursor, 27% fetch cursor 8
close cursor
– Account Rec = UOW
– No CICS tracing
– SMF 89 not collected (usage pricing)
– No Class (1) accounting
8
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
RTW Performance Data
ITR
ETR
Average ITR = 427
CPU%
Resp
ms/tran
41.49
29.3
0.032
7.06
425
62.22
42.5
0.024
6.83
439
99.59
67.5
0.016
6.78
443
245.08
176.5
0.037
7.20
417
315.65
230.7
0.081
7.31
410
V2.3
DB2 7.1
ITR
ETR
Average ITR = 426
Delta = 0.2%
9
CPU%
Resp
ms/tran
41.5
28.6
0.017
6.89
435
62.64
42.8
0.017
6.83
439
99.56
69.0
0.018
6.93
433
245.78
177.9
0.039
7.24
414
286.88
210.4
0.059
7.33
409
IBM CICS TS for z/OS V3.1 | Performance Report
V3.1
DB2 7.1
© 2003 IBM Corporation
IBM Software Group
Release to Release - Java Workloads
 Hardware
– 2084-303 system under test
– 2084-303 TPNS driver
 CICS Environment
– Single Region
– Java JDK 1.4.2
– Single JVM
– REUSE=RESET and
REUSE=YES
10
IBM CICS TS for z/OS V3.1 | Performance Report
 Software
– z/OS 1.6
 JCICS Workload
– Simple Java application
– Ave. 100 mixed API calls
– 3270 driven
 EJB workload
– Simple EJB
– Uses JCICS to call COBOL
'backend'
– COBOL application is about
3.4ms of CPU
© 2003 IBM Corporation
IBM Software Group
JCICS Performance Data
ITR
ETR
CPU%
Resp
ms/tran
ITR
ETR
CPU%
Resp
ms/tran
66.20
24.8
.009
3.74
802.1
66.26
21.3
0.007
3.21
934.5
98.57
36.6
.013
3.71
808.6
99.19
32.1
0.008
3.23
928.7
139.75
51.6
.015
3.69
813.0
141.41
44.6
0.007
3.15
952.3
193.76
71.7
.016
3.70
810.8
195.43
61.4
0.010
3.14
955.4
254.98
94.2
.092
3.69
813.0
297.27
93.6
0.037
3.14
955.4
V2.3 REUSE=YES
V2.3 REUSE=RESET
ITR
ETR
CPU%
Resp
ms/tran
CPU%
Resp
ms/tran
66.23
25.1
0.008
3.78
793.6
66.43
21.6
0.008
3.25
923.0
98.70
37.1
0.011
3.75
800.0
99.32
32.3
0.008
3.25
923.0
139.69
52.3
0.014
3.74
802.1
141.40
44.6
0.007
3.15
952.3
193.49
72.5
0.016
3.74
802.1
197.07
61.8
0.007
3.13
958.4
251.92
94.3
0.097
3.74
802.1
298.36
94.1
0.035
3.15
952.3
V3.1 REUSE=RESET
11
ITR
ETR
IBM CICS TS for z/OS V3.1 | Performance Report
V3.1 REUSE=YES
© 2003 IBM Corporation
IBM Software Group
EJB Performance Data
ITR
ETR
CPU%
Resp
ms/tran
ITR
ETR
CPU%
Resp
ms/tran
83.34
43.9
0.007
5.26
570.3
83.43
33.9
0.005
4.06
738.9
95.12
50.1
0.007
5.26
570.3
95.13
38.6
0.005
4.05
740.0
110.38
58.1
0.008
5.26
570.3
110.6
44.7
0.005
4.05
740.0
131.73
69.2
0.010
5.25
571.4
132.16
53.4
0.006
4.04
742.5
185.50
97.3
0.028
5.24
572.5
188.86
76.1
0.008
4.02
746.2
V2.3 REUSE=RESET
V2.3 REUSE=YES
ITR
ETR
CPU%
Resp
ms/tran
CPU%
Resp
ms/tran
83.32
43.4
0.007
5.19
578.0
83.16
34.5
0.007
4.14
724.6
95.12
50.4
0.007
5.29
567.1
95.19
39.4
0.006
4.13
726.3
110.38
58.5
0.009
5.29
567.1
110.49
45.7
0.006
4.13
726.3
131.73
69.7
0.011
5.29
567.1
132.15
54.6
0.006
4.13
726.3
185.21
97.8
0.029
5.28
568.1
188.79
77.7
0.009
4.11
729.9
V3.1REUSE=RESET
12
ITR
ETR
IBM CICS TS for z/OS V3.1 | Performance Report
V3.1 REUSE=YES
© 2003 IBM Corporation
IBM Software Group
HTTP
13
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
HTTP - Performance perspective
 Increase concurrent connections to 65000
– No affinity to long running CWXN transactions
 3.1 Customers with large networks
– Can now use keep-alive and SocketClose (NO)
– Now have a new CWXN for each request
– No need to make and break the connections
 Migrating from a keep-alive (no) environment
– Increased session limit and less CPU per request
 Migrating from a keep-alive (yes) environment
– Increased session limit, +3 to +4% CPU on a 2.5 MS transaction
14
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Inbound HTTP - Storage per Connection
Approx. 4.8k per connection
15
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
HTTP CPU per Tran (COMMAREA)
16
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
HTTP CPU per Tran (WEB API up to 32K)
17
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
HTTP CPU per Tran (WEB API over 32k)
Includes TCPIP
18
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
HTTPS
19
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
HTTPS - Terms
 Full handshake
– Negotiate session parameters, cipher suite etc.
– Done once per connection by CWXN for a new connection
 Partial handshake
– Done when client has previously had an SSL connection that has
since been closed
• And the Client has retained SSL session id
• The Session id has been retained in CICS storage or Coupling Facility
– Done within the CWXN transaction
 Encrption/Decryption
– For a persistent connection only the one handshake is needed
– After that just Encyption/Decryption
– Done by the CWBA transaction
20
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
HTTPS - Terms
 Hashing
– Ensure data integrity during transport (ie. not been changed),
– Two common algorithms
• MD5 Done in software
• SHA-1 Done in Hardware
 CPACF
– CP Assist for Cryptographic Function
– Available on every CP so no affinity issues
– SHA-1 TDES and DES Encypt/decrypt done by these on Z990
 PCIXCC
– Peripheral Component Extended Cryptographic Coprocessor
– Full Handshakes are done by these on Z990 if ICSF active
– Handsheke rates and card Utilisation are recorded by RMF
21
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
HTTPS - Performance perspective
 Increase concurrent connections
– No affinity to SSL TCB for duration of connection
• SSL TCB per connection now not required
– No affinity to long running CWXN transaction
 Easy choice over Cipher suites
– Specified on TCPIPSERVICE definition
 Sysplex wide cache for SSL session id
22
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
HTTPS - Storage per Connection
V3.1
Max. = 65000 connections
Approx. 10K EDSA + 3K MVS
storage per connection
23
IBM CICS TS for z/OS V3.1 | Performance Report
V2.3 - physical restriction was
17K below line per connection
+ 1.5 M above per connection
+ I TCB per connection
Max. = 250 connections
© 2003 IBM Corporation
IBM Software Group
V2.3 Vs V3.1 HTTPS
- DES 56 Encryption / Decryption costs
Data taken from RMF, includes CWXN and CWBA
24
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
HTTPS – Encryption / Decryption costs
SSL_RSA_WITH_3DES_EDE_CBC_SHA
SSL_RSA_WITH_RC4_128_SHA
SSL_RSA_EXPORT_WITH_RC2_CBC_40_MD5
SSL_RSA_WITH_DES_CBC_SHA
TLS_RSA_WITH_AES_256_CBC_SHA
25
(approx. 0.022 ms CPU per 1k)
(approx. 0.038 ms CPU per 1k)
(approx. 0.180 ms CPU per 1k)
(approx. 0.015 ms CPU per 1k)
(approx. 0.046 ms CPU per 1k)
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
HTTPS (DES 56) vs HTTP Encryption/Decryption costs
26
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
SSL Full Handshake CPU costs
Includes CSOL, CWXN and CWBA CPU costs
27
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
SSL Partial Handshake CPU costs
Includes CSOL, CWXN and CWBA CPU costs
28
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
OPEN API
29
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Open API
 Running on 'Open TCBs' gives CPU concurrency
 Throughput is not limited to the speed of a single CP
 Reduces any contention for the 'QR' TCB
 Open API is not intended to reduce CPU per
transaction
 It does give more potential throughput per region
 Allows the application to use other APIs
– Those that would otherwise 'block' the QR TCB
30
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Open API
 Prior to V3.1
– Threadsafe DB2 applications ran on L8 Open TCBs after a
DB2 call
 V3.1
– Threadsafe and CICSAPI or OPENAPI on program
definition
– OPENAPI puts program on L8 or L9 during initialisation
– USER key = L9 CICS key = L8
31
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Open API
CICS API
OPEN API
CPU secs used per minute
at 482 trans per sec
CPU secs used per minute
at 1065 trans per sec
Quasi - max. ETR = 482
CPU per tran = 2.26ms
OPEN - max. ETR = 1065
CPU per tran = 2.54ms
32
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Open API Cautions!
 Choose applications carefully
– Avoid applications with non-threadsafe APIs
• e.g.. lots of File Control (this will cause TCB switching back to QR)
 Migrating Threadsafe DB2 applications from V2.3
– userkey/threadsafe on V2.3 ------------------- application runs on L8
– userkey/threadsafe/OPENAPI on V3.1 ----- application runs on L9
• DB2 calls then need to switch to L8!
– Leave DB2 Threadsafe applications moving from V2.3 as
CICSAPI!
33
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
XPLINK
34
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
XPLINK
 XPLINK introduced in OS/390 V2.10
– feature of z/OS that provides high performance subroutine call and
return mechanism
– Supported by C/C++ compiler of z/OS
– Specified by C/C++ compiler option XPLINK
– Previously not supported in CICS
 XPLINK requires MVS LE rather than CICSLE
– Therefore application requires own TCB to run on
 XPLINK programs execute like OPENAPI programs
– Runs on an X8 or X9 TCB with MVS/LE rather than L8 or L9 with
CICS/LE
– CICS detects at runtime that program compiled with XPLINK
35
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
XPLINK cautions!
 XPLINK programs run on X8 or X9 TCBs
– Mixing DB2 calls will cause switches to L8 TCBs
– Might be better not using XPLINK but C++ with
THREADSAFE CICSAPI for DB2
 Mixing lots of non_threadsafe CICS API will cause
excessive switching to QR TCB
 Foundation Classes are currently not XPLINK
compiled
– A call to one of these will cause a TCB switch
36
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Channels and
Containers
37
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Performance Evaluation Configuration
 COBOL program links to COBOL program
 2.5ms of application program logic
 Comparison of data transfer using:
– Commareas
– Channels and Containers
– Temporary Storage Queues
38
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Commarea application
Sender program
Receiver program
EXEC CICS GETMAIN
EXEC CICS LINK ...
EXEC CICS RETURN
COMMAREA
EXEC CICS RETURN
32K max data restriction
39
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Channel application
Sender program
EXEC CICS GETMAIN
EXEC CICS PUT
CONTAINER
EXEC CICS LINK ...
Receiver program
EXEC CICS GET
CONTAINER
EXEC CICS RETURN
CHANNEL
EXEC CICS RETURN
Single container used for data
No restriction on data size
40
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
TSQ application
Sender program
EXEC CICS GETMAIN
Receiver program
EXEC CICS READQ TS
EXEC CICS WRITEQ TS
EXEC CICS RETURN
EXEC CICS LINK
EXEC CICS DELETEQ TS
EXEC CICS RETURN
Multiple TSQ items used for data size >32K
41
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Single Region Links < 32K
42
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Single Region Links >32K
43
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
MRO Link < 32K
44
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
SOAP Pipeline
45
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
SOAP Provider Pipeline
HTTP
SOAP
Client
SOAP Pipeline
Handles
HTTP layer
and SOAP
Envelope
46
IBM CICS TS for z/OS V3.1 | Performance Report
App Handler Business Logic
Business
Converter
between
logic program
XML SOAP
Body and
data structure
© 2003 IBM Corporation
IBM Software Group
SOAP support in CICS
 CICS SOAP support
– SOAP for CICS Feature in V2.2 and V2.3
• HTTP or WebSphere MQ used at transport layer
• user-written application handler required
• commarea-based business logic application
– SOAP support integrated into V3.1
• HTTP or WebSphere MQ used at transport layer
• user-written application handler NOT required
• business logic application can be:
– commarea or container based
47
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Application Handler
 V2.2 and V2.3 options
– User-written application required to:
• Convert XML SOAP body input to language data structure
• Invoke commarea-based business logic program
• Convert language data structure output to XML SOAP body
– WSED converter (WSED version 5.1.2)
• WebSphere Studio Enterprise Developer
• Generates driver, input and output programs from business logic
program source code
• Upload programs to z/OS and install in CICS
– New Handler required for each business logic program
48
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
CICS Web Services Assistant
 Batch component
– DFHLS2WS batch program
• Language data structure input
• HFS WSDL and WSbind files created
– DFHWS2LS batch program
• HFS WSDL file input
• Language data structure and HFS WSbind files created
– New WSbind file required for each business logic program
49
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
CICS Web Services Assistant
 Runtime component
– Pipeline install process
• Accesses WSbind files from pickup directory
• Copies WSbind files to shelf directory
– Runtime
•
•
•
•
50
Pipeline reads WSbind file
Converts input XML SOAP body to language data structure
Invokes commarea/container based business logic program
Converts output language data structure to XML SOAP body
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Performance Evaluation #1
 Compare for 'simple' SOAP message:
– SOAP for CICS Feature in V2.3
• with user-written app handler
– SOAP pipeline in V3.1
• with user-written app handler
– SOAP pipeline in V3.
• using CWSA (CICS Web Services Assistant)
51
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Simple SOAP Message
msecs
cpu per
transaction
52
V2.3
User
Program
1.35 ms
V3.1
User
Program
1.13 ms
IBM CICS TS for z/OS V3.1 | Performance Report
V3.1
CWSA
without
PTF
1.28 ms
V3.1
CWSA
with
PTF
1.11 ms
© 2003 IBM Corporation
IBM Software Group
Performance Evaluation #2
 Compare for 'typical' SOAP message:
– SOAP for CICS Feature in V2.3
• with user-written app handler
– SOAP pipeline in V3.1
• with user-written app handler
– SOAP pipeline in V3.1
• using CWSA
– SOAP pipeline in V3.1
• with WSED converter app handler
53
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Typical SOAP message
 The next series of evaluations all use the 'typical'
SOAP message
– These include:
• SOAP provider and requester environments
• HTTP and WebSphere MQ transport layers
 Variations in SOAP message size and # of elements
are discussed later
54
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Typical SOAP Message
msecs
cpu per
tran
55
V2.3
User
Program
3.41 ms
V3.1
User
Program
3.46 ms
IBM CICS TS for z/OS V3.1 | Performance Report
V3.1
WSED
Converte
r
6.47 ms
V3.1
CWSA
without
PTF
4.47 ms
V3.1
CWSA
with PTF
2.64 ms
© 2003 IBM Corporation
IBM Software Group
Typical SOAP Message
SOAP2.3 comparison:
WSED3.1
+89.7%
V3.1
-1.5%
CWSA3.1
-22.6%
56
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Performance Evaluation #3
 Compare for 'typical' SOAP message:
– CICS SOAP requester region
• SOAP for CICS Feature V2.3
• SOAP pipeline in V3.1 using CWSA
– CICS SOAP provider region
• SOAP for CICS Feature V2.3
• SOAP pipeline in V3.1 using CWSA
– Transport Layer communication
• HTTP
• WebSphere MQ (WMQ)
57
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
CICS Requester and Provider Environment
Http or WMQ
3270
screen
input
Requester Application
Receive
Put Container
Invoke Webservice
Get Container
Send
Return
Provider Application
Handle input commarea
Generate output commarea
Return
<========== LPAR ==========>
58
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
SOAP HTTP Requester - V3.1 vs V2.3
CWSA HTTP
59
-19%
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
SOAP Requester - HTTP vs WMQ
CWSA WMQ
60
+53.6%
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
SOAP Provider - HTTP vs WMQ
CWSA WMQ
61
+30.5%
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Variations
The table below shows how the CPU usage changes with
variations in the processing of the SOAP message
Change:
% cpu increase
Commarea to Container program
0.91%
SOAP1.1 node to SOAP1.2 node
0%
Add header processing program
3.95%
Note: Header processing programs within the SOAP
node are invoked on input and on output
62
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
SOAP inbound message size
1.1ms for null SOAP msg
0.05ms for every K of data
63
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
SOAP outbound message size
1.1ms for null SOAP msg
0.042ms for every K of data
64
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
SOAP inbound elements
1.1ms for null SOAP msg
0.043ms for every 100 elements
65
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
SOAP outbound elements
66
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Sizing SOAP message
SOAP message consists of:
SOAP Prefix and Suffix
<?xml version='1.0' encoding='UTF-8'?>
<SOAP-ENV:Envelope xmlns: .............>
<SOAP-ENV:Body>
</SOAP-ENV:Body>
</SOAP-ENV:Envelope>
Operation name
<PGRMNAMEOperation>
</PGRMNAMEOperation>
SOAP Body elements
<element1>1234</element1>
67
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Sizing SOAP message
SOAP message size is the sum of:
 SOAP Prefix and Suffix
– 278 bytes
 Operation name
– (program_name_length*2)+23 bytes
 SOAP Body elements
– (avg_element_name_length*2)+avg_element_data_length+5
– multiply by number of elements
68
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Sizing Example
Example with:
4K commarea
program name 8 bytes
average element name length 8 bytes
# elements
128
512
2048
4096
69
average
SOAP msg
data length size (bytes)
32
7,101
8
15,165
2
47,421
1
90,429
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Sizing Example
Example with:
4K commarea
program name 8 bytes
# elements
128
512
2048
4096
70
avg len of
element
8
16
24
30
IBM CICS TS for z/OS V3.1 | Performance Report
average
SOAP msg
data length size (bytes)
32
7,101
8
23,357
2
112,957
1
270,653
© 2003 IBM Corporation
IBM Software Group
Codepage conversion
71
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Performance evaluation overview
 z/OS Conversion Services
 Evaluations performed within CICS using containers
 Conversion occurs between PUT and GET container
 Codepages used in testing
– EBCDIC
(037)
– ASCII
(437)
– UTF-8
(1208)
– UTF-16
(1200)
 2 stage conversions for UTF-8 to/from EBCDIC or
ASCII
72
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
Results summary
 CPU milliseconds to convert 1K characters
– UTF-8
(1208) to ASCII
(437)
– UTF-8
(1208) to EBCDIC
(037)
– UTF-8
(1208) to UTF-16
(1200)
– ASCII
(437)
to UTF-8
(1208)
– EBCDIC
(037)
to UTF-8
(1208)
– UTF-16
(1200) to UTF-8
(1208)
– EBCDIC
(037)
to UTF-16
(1200)
– ASCII
(437)
to UTF-16
(1200)
– UTF-16
(1200) to EBCDIC
(037)
– UTF-16
(1200) to ASCII
(437)
– EBCDIC
(037)
to ASCII
(437)
– ASCII
(437)
to EBCDIC
(037)
73
IBM CICS TS for z/OS V3.1 | Performance Report
0.0094
0.0093
0.0090
0.0087
0.0087
0.0076
0.0031
0.0030
0.0021
0.0021
0.0014
0.0014
© 2003 IBM Corporation
IBM Software Group
'WRAPPERS'
74
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation
IBM Software Group
'V3.1 Wrappers‘ Relative CPU costs
Data taken from RMF
75
IBM CICS TS for z/OS V3.1 | Performance Report
© 2003 IBM Corporation