FW: z/OS V1R8 WLM CICS Performance Observation
Un-related to new z/OS release - I would check relative MVS priority of CICS main TCB (QR) where the CICS dispatching is done - to be higher that other L8/L9/etc.. TCBs in CICS address space.. Shmuel Koller From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Jim Marshall Sent: Tuesday, February 27, 2007 10:38 PM To: IBM-MAIN@BAMA.UA.EDU Subject: z/OS V1R8 WLM CICS Performance Observation We run CICS TS ADABAS and saw some performance issues with CICS transactions in z/OS V1R7 although not very bad. When we brought up z/OS V1R8 we are seeing high level of delays in CICS transactions when they are going to redispatch queue. Talking to IBM lab folks they are telling us it is not a z/OS V1R8 problem but maybe a CICS performance problem. But then the CICS people say the transactions are not being dispatched quick enough off the redispatch queue. I think this has been shuffled over to the WLM folks. Like to know if anyone else is encountering CICS performance problems where it seems the CPU is running 100%, transactions start queueing up, repsonse gets slow, CPU drops down to 50-60%, redispatch queue gets bigger and then all them transactions come out at once and FLY through the system. Is this happening to you or has it happened lately. thanks Jim -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: z/OS V1R8 WLM CICS Performance Observation
We got rid of ADABAS before WLM, but isn't it a separate started task? I don't think the ADABAS work done on behalf of CICS ran under the CICS TCB's (like DB2 IMS does) when we had it, it may have changed. If not, you have to find some balance between the priorities of CICS vs. ADABAS tasks. -Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Shmuel Koller Sent: Thursday, March 01, 2007 4:10 AM To: IBM-MAIN@BAMA.UA.EDU Subject: FW: z/OS V1R8 WLM CICS Performance Observation Un-related to new z/OS release - I would check relative MVS priority of CICS main TCB (QR) where the CICS dispatching is done - to be higher that other L8/L9/etc.. TCBs in CICS address space.. Shmuel Koller -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
z/OS V1R8 WLM CICS Performance Observation
We run CICS TS ADABAS and saw some performance issues with CICS transactions in z/OS V1R7 although not very bad. When we brought up z/OS V1R8 we are seeing high level of delays in CICS transactions when they are going to redispatch queue. Talking to IBM lab folks they are telling us it is not a z/OS V1R8 problem but maybe a CICS performance problem. But then the CICS people say the transactions are not being dispatched quick enough off the redispatch queue. I think this has been shuffled over to the WLM folks. Like to know if anyone else is encountering CICS performance problems where it seems the CPU is running 100%, transactions start queueing up, repsonse gets slow, CPU drops down to 50-60%, redispatch queue gets bigger and then all them transactions come out at once and FLY through the system. Is this happening to you or has it happened lately. thanks Jim -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: z/OS V1R8 WLM CICS Performance Observation
Jim, There's not enough information yet. Are you running CICS with response time goals? Are they being met? If you are using velocity goals only, are those goals being met both before and after the queue buildup? Input! Input! Steve Samson Jim Marshall wrote: We run CICS TS ADABAS and saw some performance issues with CICS transactions in z/OS V1R7 although not very bad. When we brought up z/OS V1R8 we are seeing high level of delays in CICS transactions when they are going to redispatch queue. Talking to IBM lab folks they are telling us it is not a z/OS V1R8 problem but maybe a CICS performance problem. But then the CICS people say the transactions are not being dispatched quick enough off the redispatch queue. I think this has been shuffled over to the WLM folks. Like to know if anyone else is encountering CICS performance problems where it seems the CPU is running 100%, transactions start queueing up, repsonse gets slow, CPU drops down to 50-60%, redispatch queue gets bigger and then all them transactions come out at once and FLY through the system. Is this happening to you or has it happened lately. thanks Jim -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: z/OS V1R8 WLM CICS Performance Observation
I agree with Steve (a typical occurrence), more info is needed. At z/OS 1.7, did you routinely see your processor at 100% busy? What about the other LPARs on the box (if there are any)? Weights properly set? The reason for this line of questioning: if you were routinely running just under 100% at z/OS 1.7, when you went to 1.8, your overhead went up 2% to 3% to 5% (your mileage will vary). Once the CPU busy gets to 100%, WLM will now kick in. Once the entire CEC goes to 100% busy, PR/SM will kick in. An non-optimal WLM Service Policy along with improper LPAR controls may be tolerated at 95% busy; but once over 100% may cause you pain. Are you running on a uni-Processor? This also has challenges. So other input needed: Physical Config, Logical Config, WLM Service Policy (helpful), Perf Index for problem Service Class Period(s), what else is running in the system (above your problem CICS) Without more info, I'd suggest it is NOT a CICS problem (unless you also upgraded CICS or added maintenance). -Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Steve Samson Sent: Tuesday, February 27, 2007 SYSN PM 1:31 To: IBM-MAIN@BAMA.UA.EDU Subject: Re: z/OS V1R8 WLM CICS Performance Observation Jim, There's not enough information yet. Are you running CICS with response time goals? Are they being met? If you are using velocity goals only, are those goals being met both before and after the queue buildup? Input! Input! Steve Samson -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html