Hi Chris and Community,
Point taken on getting closer to the microphone.
To start the counting of SWAP you can specify the following in SYSIDMS
CSATST49
Yes. The SWAPs out of RHDCCXIT are for the user defined exits in RHDCUXIT. The SWAPs out of IDMSDBMS are for the DB Procedures. The SWAPs in OCEP2 are for physical I/Os.
If you don’t mind, please open an issue for the EXIT 29 request. We can certainly look into a modification in this area. I will have to run this up the flag pole here but doesn’t seem unreasonable. You could also open an issue related to the DB procedure.
With respect to the questions on what the displays are telling you.
For the “JOB OUTPUT”. The “TCB” CPU time is the true TCB CPU time used by the job. This value should match what SMF records and should be very close to what the DCMT commands display. (although your displays below do not reflect this. I think you may be missing a PTF or there is another bug in this area. An issue should be opened for this mismatch). Again in the “JOB OUTPUT”, the SRB CPU time is not related to the “zIIP” SRB CPU time. This value represents the time spent handling “SRBs” scheduled into the IDMS address space.
Due to the different in CPU time being reported as noted above I’m not able to comment on the CPU effectiveness since the values may be incorrect. I am used to seeing the swap attempted and actual swaps being fairly close. There is a pretty big difference in your display below. If you want you can post the entire DCMT D SUBT display showing all the locations where the swaps are occurring along with a DCMT D SUBT 1 issued at the same time. A “DCMT D STAT SYS” would be a good display to do roughly at the same time as well. You can also just include these in the issue related to the times being suspect as noted above.
John Siraco
CA Technologies
Director, Software Engineering
Tel: +1 508 628 8247
John.Siraco@ca.com
<mailto:
John.Siraco@ca.com>[Description: C:\Program Files\CA\GIS\CASig\CA_email.gif]<http://www.ca.com/>
From: CA IDMS IUA EIUA Global User CommunityMessage Boards [mailto:
CommunityAdmin@communities-mail.ca.com]
Sent: Wednesday, November 27, 2013 10:01 AM
To:
mb_message.15281481.105393767@myca-email.ca.com
Subject: [DBA Issues - Open Forum] RE: Webcast Replay: CA IDMS gets even zIIP-ier
Hi – I am Chris Trayler. IDMS DBA at Bank Julius Baer in Zurich.
I managed to watch the Webcast this morning. It was very informative and worth listening to. I have a few comments and questions.
First of all, well done Iain and Chris for stepping up to the plate on behalf of the community.
Please can you wheel John closer to the microphone next time. We IDMS DBAs are all getting to the age where our hearing is going.
I have applied the APARs to our test systems for evaluation. This morning I also followed Iain’s suggestion and turned CSATST49 on. He wasn’t quite accurate in his description. The SYSIDMS parm is CSATST49. Coding CSATST49=ON as he said results in a startup eror.
To turn it on dynamically the flag is at offset 4BA in the CSA. It is very easy to use DCMT VARY MEM to activate it. That also has the advantage of missing out all the swaps at startup which you don’t really want to know about.
We are procedure and exit heavy here. It seems there is some scope for improving it.
From the display enabled by CSATST49 I surmise that RHDCCXIT swaps are from our Exits and that IDMSDBMS swaps are from the DB procedures.
Iain alluded to “coding changes” when he referred to Exit 14. We have Tracer from ASG which has an Exit 14. That Exit is overwhelmingly our biggest issue. Can Iain or ASG share any tips?
I also have an Exit29. This is a bank with a high emphasis on monitoring access – especially via ad hoc SQL queries. We use Exit 29 in conjunction with the relatively new SAVE_SQL_SYNTAX=ON SYSIDMS parm to report on SQL queries. CA could do me a big favour with Exit 29.
Exit 29 is the Security check exit which is called for every security check in the system. I am only interested in one type of Security check which is an SQL Schema Check.
SCHECK CLI SRBFUNC,X'0C' IS IT AN SQL SCHEMA CHECK
BNE RETURN
Those calls are relatively rare in the grand scheme of things. 99% of the time in Exit 29 I do nothing – but I have been called and so I get a swap. If I could get an optional APAR to invoke Exit 29 only on an SQL Schema check then I could eliminate several million unnecessary swaps.
I also have a frequently called DB Procedure – again it is monitoring/security based. I implement a form of row-level security for particularly sensitive records. This is only relevant to our Test systems which have a wide range of users. In production the system is heavily secured in other ways and the DB procedure is not required. But the Schemas are identical between Test and Production. In Prod I implemented a dummy procedure which is basically an IEFBR14. That has worked fine with a minimal overhead for many pre zIIP years. Now I will get a Swap for each invocation of the dummy proc. Obviously I can complicate our system a bit by eliminating the DB Proc call from the production Schema. I didn’t really want that. I like Test and Prod to be the same. Other procedures like IDMSDCOM and IDMSCOMP can be called without a swap. If CA could supply a do-nothing non-swapping proc – with the possibility of a named entry point or an alias or a rename – then I could get round it without a different Schema between Test and Prod.
I am a little confused about what DCMT D SUBT is telling me. Here is an example of a before and after APAR 18.0 display. It is a single tasking CV doing mainly batch and a little CICS work. Work load was similar on both days.
Special concerns – CPU Effectiveness? Swaps? Comments?
Note that the real tape number is GJI00B. I actually zap the necessary modules to do my “notional” service packs like in the old days. GJI04M and GJI05M don’t really exist.
Before zIIP APAR
From Job output
16.16.21 JOB02333 -JOBNAME STEPNAME PROCSTEP RC EXCP CPU SRB CLOCK SERV PG PAGE SWAP VIO SWAPS STEPNO
20.24.52 JOB02333 -TIDMS12I IDMS#129 IDMSEXEC 00 33301K 42.89 6.65 1688.5 803747 0 0 0 0 0 3
DISPLAY SUBTASK EFFECTIVENESS
*** Subtask display ***
Subtask Elapsed time Total CPU time % CPU SRB
Name TCB SRB TCB SRB TCB SRB
-------- -------------- -------------- -------------- -------------- --- ---
MAINTASK 48:28.112206 01:07:28.9866 41:23.095777 41:23.095571 85 61 Y
-------- -------------- -------------- -------------- -------------- --- ---
Totals 48:28.112206 01:07:28.9866 41:23.095777 41:23.095571 85 61
V129 ENTER NEXT TASK CODE: CA IDMS release 18.0 tape GJI04M node TIDMS12I
*** Display Subtask details ***
Name MAINTASK
Number 01
Status BUSY
Work type IDMS
Count wakeups 71,506,084
Count task dispatches 253,656,788
User mode CPU time 00:00:00.036942
System mode CPU time 01:23:02.315339
CPU effectiveness (%) 85
Count times fast posted 103,788,443
Count times OS posted 00
Count found work pass 1 252,106,051
Count found work pass 2 1,550,737
Count times POSTEXIT resumed 71,380,136
*** Enclave Info ***
zIIP time 00:41:22.720400
zIIP on CP time 00:00:08.455612
CPU effectiveness (%) 61
Count swap attempts 248,627,417
Count actual swaps 163,516,008
V129 ENTER NEXT TASK CODE: CA IDMS release 18.0 tape GJI04M node TIDMS12I
After zIIP APAR
From Job output
06.00.14 JOB03243 -JOBNAME STEPNAME PROCSTEP RC EXCP CPU SRB CLOCK SERV PG PAGE SWAP VIO SWAPS STEPNO
20.24.27 JOB03243 -TIDMS12I IDMS#129 GJR18SP5 00 29994K 20.50 7.21 864.21 313810 0 0 0 0 0 2
D SUBT EFF
*** Subtask display ***
Subtask Elapsed time Total CPU time % CPU SRB
Name TCB SRB TCB SRB TCB SRB
-------- -------------- -------------- -------------- -------------- --- ---
MAINTASK 01:38:18.5595 02:39:21.7150 20:07.090044 54:29.488890 20 34 Y
-------- -------------- -------------- -------------- -------------- --- ---
Totals 01:38:18.5595 02:39:21.7150 20:07.090044 54:29.488890 20 34
V129 ENTER NEXT TASK CODE: CA IDMS release 18.0 tape GJI05M node TIDMS12I
D SUBT 001
*** Display Subtask details ***
Name MAINTASK
Number 01
Status BUSY
Work type IDMS
Count wakeups 66,753,529
Count task dispatches 251,679,692
User mode CPU time 00:00:00.024386
System mode CPU time 01:14:36.573694
CPU effectiveness (%) 20
Count times fast posted 103,578,410
Count times OS posted 00
Count found work pass 1 249,845,493
Count found work pass 2 1,834,199
Count times POSTEXIT resumed 66,695,206
*** Enclave Info ***
zIIP time 00:54:24.769060
zIIP on CP time 00:00:04.737393
CPU effectiveness (%) 34
Count swap attempts 201,458,590
Count actual swaps 201,404,908
V129 ENTER NEXT TASK CODE: CA IDMS release 18.0 tape GJI05M node TIDMS12I
Posted by:christopher.trayler
--
CA Communities Message Boards
105396307
mb_message.15281481.105393767@myca-email.ca.com<mailto:
mb_message.15281481.105393767@myca-email.ca.com>
https://communities.ca.com