| |
|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
  |
|
| CICSREQUEST node performance |
« View previous topic :: View next topic » |
| Author |
Message
|
| PEPERO |
 Posted: Tue Oct 27, 2015 9:38 am Post subject: CICSREQUEST node performance |
 |
|
Disciple
Joined: 30 May 2011
Posts: 177
|
Hi all;
Does CICSSVC value in EXCI options table has any performance impact on the CICSREQUEST node throughput for IBM WMB V7.0.0.1 or higher in which it uses IPIC over TCP/IP? We have CICSTS4.1 under Z/OS 1.11. |
|
| Back to top |
|
 |
| pmasters |
| Posted: Wed Oct 28, 2015 2:25 am Post subject: |
|
|
Novice
Joined: 19 Jul 2011
Posts: 14
|
Hi Pepero,
So if you're using IPIC, then the CICSSVC value won't affect you. As you say, the SVC is used by EXCI.
Performance options which will affect IPIC are things like the number of receive sessions on the link (how many parallel tasks can be in progress). Note also that any kind of of transactional work will mean that the session used by the CICSRequest node will be retained until transaction commit/rollback so you may end up using more of them, so check its not bottlenecking.
One further consideration is to see whether you can exploit use of the threadsafe mirror transaction to use an Open TCB on the CICS side, otherwise you will share the QR TCB.
Thanks, Peter
_________________
===========
Peter Masters |
|
| Back to top |
|
|
| PEPERO |
| Posted: Wed Oct 28, 2015 8:41 am Post subject: |
|
|
Disciple
Joined: 30 May 2011
Posts: 177
|
Although we are using dynamic IPCONN cics resources in which a default of 100 simaltanous conversations is supported, having the CICS statistics report shows that the flow never reaches to that limit and it reports 15~20 concurrent sessions are active. When turning on the WMB statistics report it shows a considerable waste time on the CICSREQUEST nodes.
In fact the CICS application program that is target in the test scenario is as simple as possible and doesn't have any IOs.
When we use pure CTG (client <=> CTG <=> CICS) to make access to the CICS application server , it has a good performance (~ over 1400 tps) . Of course i've to say we have set the CTG to use EXCI interfaces to connect to CICS instead of having IPIC connection.
I'd found an IBM UK LAB documentation in which a performance test was done between CICSREQUEST node and CICS MQ bridge in an Z900 IBM hardware with the FICON ESS model 800 DASD with 4-CPU LPAR on which Z/OS 1.4 , CICS TS 2.2 , WBIMB 5.01 and WMQ 5.3.1 was installed. The document shows a message rate and CPU cost for 3 scenarios(CICS and WMB don't use global transaction and are resided in a single LPAR , CICS and WMB don't use global transaction and are resided in separated LPARs , CICS and WMB participate in a global transaction). The conclusion was when using separte transactions between CICS and WMB with 4 message flow threads a total tps of 1600 is achived using CICSREQUEST node , three times faster message processing rate and two times faster on CPU consumption. We have ~ 500 tps with 4 threads. At this document the CICSREQUEST node version used , doesn't support IPIC connection and they defined 4 EXCI session terminals to connect to CICS. Since our pure CTG test scenario has an acceptable throughput for the same CICS application program , i think some thing must be wrong for IPIC connection definition between WMB and CICS. |
|
| Back to top |
|
|
|
|
|
| |
|
Page 1 of 1 |
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
|
|
