Chairman: Frank Fleming frank.fleming@barclays.co.uk 01565-613581
Secretary: John Boyle john_boyle@bmc.com 01784-478832
Web Site: http://web.ukonline.co.uk/dougie.lawson/gse/
Printed copies of some of the presentations were distributed at the meeting. If anyone wants to receive a copy of any of the presentations please email working group chairman - Frank Fleming. A copy of the "IBM What's New" presentation will be posted to the Web Site.
A gift was presented to Andrew Jack (Standard Life) in recognition of his contribution to the Working Group over the years, in his capacity as membership secretary.
Dougie was standing in for Niel Kenyon. A copy of the presentation is posted on the IMS Working Group web site.
In addition to education offered in the UK, there are a number of PSSC courses available which may be of interest, including
These courses are offered in Montpellier, France and further details can be found on the French Section of the IBM Learning Services Web Site. (Course description etc. is in English!)
Dougie covered the following IMS and IMS Related Product Announcements.
NOTE that Service for IMS/ESA Version 5 will be discontinued from 30th September 2001.
IMS V7 Promotion - Complimentary IMS Migration Assistance Package is being offered to all customers who order IMS Version 7 before October 15, 2001. This includes on day on-site visit by IMS Specialist.
See presentation for details of recent IMS Redbooks
John's presentation covered new initiatives and product updates from BMC
Further details from BMC's web site at http://www.bmc.com/
Aurora's presentation covered coming enhancements to Candle's IMS Products, and coming UK Events.
Brian gave us an interesting overview of the Tivoli Business Systems Manager. This product provides an overall, top-down view of total state of your environment. The product is repository based, and the size of the repository can be considerable in a large enterprise. The product relies on auto-discovery of objects, using OS/390 batch jobs to populate the repository. Each delivery is customized to the specific environment, and can be adapted to utilise existing management tools. The system view is provided via a Graphical User Interface (GUI) running on UNIX/NT.The GUI can be set up in 'Blind Mode' where ICONs only appear if there is a problem.
In the IMS environment the product uses the AOI Exit, then transmits information to its server using LU6.2. As delivered it operates in 'display only' mode, but can with existing tools like TMON, OMEGAMON, MAINVIEW.
More information available form Brian, who will be glad to arrange a site visit to see the product in operation. Or on the we at http://www.tivoli.com/products/index/business_system_mgr/
Joe's presentation covered Standard Life's migration from IMS 5.1 to 7.1, which took place between February and June of this year. He explained that the main driver behind the upgrade was that IMS 5.1 would be 'out of support' at the end of September. As part of the migration they took the opportunity to remove an in-house written routine which acted as a front-end to the IMS Region Controller. This avoided updating the routine, which read the RECONs direct. Some of the function of the front-end routine (Parm verification) was moved to a JES exit.
Environment - 2 Development systems, 3 Production systems, IMS Full Function and DB2. 3 to 4 Million transactions per day. MQ.MQSI/OTMA to connect Distributed Systems Applications to IMS, plus CICS Transaction Gateway.
Issues - Some issues with Vendor Support - although IMS 7.1 was GA, some vendors were not sure of the release level required for IMS 7.1.
Customization - One security related problem with an MSC Exit.
Code Problems - No significant problems - some minor.
Application Issues - One 'legacy' application which uses OTMA receiving DFS2082 RESPONSE MODE TRANSACTION TERMINATED WITHOUT REPLY. Application worked OK in IMS 5.1. Temporarily circumvented by local Usermod to IMS, to allow time for application changes.
During Joe's presentation on their V7 migration experience he mentioned that their OTMA applications, now generate DFS2082I messages, which they didn't use to with IMS V5 - this is for programs doing pgm-to-pgm switches.
They were wondering of it would be possible to have the same capability for OTMA that APPC has with APAR PQ45320 - so that the DFS2082I message doesn't get sent back to the client from one of the programs in the pgm-to-pgm switch sequence, as the last program that is invoked does the insert to the originating client.
I contacted the OTMA developer and he is planning on providing the similar option to APPCASY for OTMA for IMS V7 (OTMAASY) by 1Q2002.
Conclusion - Code was stable but a lot of HIPERs to chase around. This was time consuming, even utilizing Advanced Hold Data.
Future - HALDB could be considered to replace user written partitioning. Will try to avoid skipping IMS Releases in future.
Frank gave an overview of Barclays' participation in the IMS 7.1 QPP. He stated that Barclays' main reasons for participating in the QPP were to review and influence the release content, and to identify compatibility issues. Unlike the IMS 6.1 QPP, where there was a need to get to production as soon as possible to support Data Sharing for DEDBs, there was no immediate business driver, for IMS 7.1.
Main Features Tested - Base code, Utilities, Usermods. MSC Exit consolidation - DFSMSCE0. HALDB, using own test bed rather than just the IVP. Online Recovery Service (ORS) - now packaged as a separate product. Most of the other main line items.
Issues - Raised a total of 42 PMRs. Required PUT 0101C for BMC's MAINVIEW for IMS.
Live/Production Implementation - delayed until February 2002 due to various contractual issues.
| Name: | Neil Price |
| Company: | TNT |
| Problem | reported that they went live on IMS 7.1 DBCTL during July. The implementation
went smoothly. His only complaint was the number of unresolved PEs. Since
then they had hit one bug which caused a system outage.
The problem was initially triggered by a CICS application program which caused a CICS storage violation and the CICS Region abended leaving an in-doubt thread. (This was only possible because the transaction was erroneously defined to run in the File Owning Region; this used to be normal for TNT's background transactions, but was changed recently to avoid exactly this type of problem.) The CICS Region was restarted, and IMS resolved the in-doubt thread, but as a result of an IMS bug managed to corrupt the PST. This bug is described in HIPER APARs PQ48213 (for V7) and PQ47593 (for V6). The PTFs are UQ54151 and UQ54150 respectively. An IMS cold start and/or reformat of the RDS is required to clear the problem. Because of the corruption of the PST (two PSTs finish up with the same region id), and any subsequent transactions to use that region also abended. The corrupted PST caused MAINVIEW to abend U3000-02, and subsequently back itself out of the DBCTL control region. Unfortunately this exposed a bug in MAINVIEW which caused the DC MONITOR to be started when MAINVIEW backed out. This problem was notified to MAINVIEW customers by a aTechnical Bulletin on August 8th. The PTFs are BPI8328 (for V3.3.) and BPI8070 (for V3.2). Running with DC monitor on aused excessive CPU usage by BMPs, and a gradual degradation of performance until CICS was also affected by locks being held by BMPs.
|
| Name: | Phil Courtney |
| Company: | Bank of America |
| Problem: | commented that he was having problems setting up RRS (OS/390 Resource Recovery Services). He is unsure of what information to enter on the ISPF dialogues. |
| Answer: | Alison Coughtrie (IBM) offered to supply the name/number of a Red Book
on this subject. Dougie Lawson (IBM) can provide a ZAP to disable the service
on IMS Systems where it is not required.
|
| Post meeting update from Alison Coughtrie | 1. From IMS V6 onwards, IMS will register with RRS when it initializes.
Some customers have noticed an increase in IMS response times when RRS is
active on the same OS/390 as IMS (even though they are not using protected
conversations in IMS). This is due to the fact that IMS registers unprotected
interest with RRS for each 3270-based tran, and that there is an RRS ARCHIVE
LOG entry for each of these completed URs.
2. For those customers who are not using protected conversations in IMS (but are using RRS for something else), some have bypassed the problem by applying a ZAP which prevents IMS registering with RRS. (PMR 61949, B215, C616 has the details of the ZAP). This PMR has however been updated recently - this customer disabled RRS archive logging and was then happy with the performance and got rid of the ZAP. 3. OW47139 is an RRS APAR to address IMS transaction times when using the ARCHIVE log. It also gives recommendations as to how the log should be set up. Apparantly this APAR will "Make the archive log write process run under a server task in the RRS address space. This will make the write to the archive log run asynchronously with respect to the IMS transaction, so the IMS transaction will not have to wait." 4. It seems that if you want to implement protected conversations in IMS, you should either not use the RRS ARCHIVE log, or make sure you have OW47139 on your system if you do. That should lessen the affect on the non-protected transactions, and the overheads then would depend on what the protected conversations are doing and the volumes. It has been recommended not to use the Archive Log even with OW47139 applied, because you have to do something with the data. Most customers have absolutely no use for it and if the log stream fills up....
|
| Name: | Neil Price |
| Company: | TNT |
| Problem: | would like to hear from anyone who has experience of moving from PI to
IRLM for IMS lock management. They intend to go to Data Sharing, but want
to implement a 'Single Member Sharing Group' under IRLM first, before 2-way
data sharing.
|
| Answer: | No offers. |
| Name: | Phil Courtney |
| Company: | Bank of America |
| Problem: | Asked if anyone had any up to date mappings for IMS log records, to assist
in scanning IMS Logs for information. He used to use SAS/MXG to do this.
|
| Answer: | No offers, but a suggestion to write a DFSERA10 exit to accomplish
this. Alternatively - try IMS List Server for offers.
|
| Name: | Phil Courtney |
| Company: | Bank of America |
| Problem: | Asked if those who had moved to IMS 7.1 had seen any performance improvement.
|
| Answer: | Joe Owens (Standard Life) said that they saw no change (IMS/TM 5.1 to
7.1 upgrade). Neil Price (TNT) said the same (DBCTL 6.1.to 7.1 upgrade). Frank Fleming (Barclays) said that they saw a reduction in cpu when they moved from IMS 5.1 to 6.1. Dougie Lawson (IBM) reported that an IMS 7.1 Performance Red Book was currently being written.
|
Alison opened by stating that there were 18 sites with IMS 7.1 in production at the GA date. Today's presentation provided a very high level summary of the main items included in IMS 7.1, and a more detailed look at the following items which have been delivered via the Service Process, since 7.1 went GA, or which are planned for delivery.
Already Delivered
Additional enhancements planned
Alison also mentioned the IMS V7 promotion to assist in migration from IMS 5.1 or 6.1
As part of the planning process for future releases, Alison would like to know if any Fast Path users still have V5 format SDEP CIs on their DEDBs. None of the sites present did - but if anyone else has, please let Alison know.
In my presentation I said that Fast Path SVSO Duplexing was not supported in a GDPS. This restriction no longer applies. Fast Path SVSO duplexing and SVSO system-managed duplexing are both supported in a Geographically Dispersed Parallel Sysplex environment.
The discussion topic has been active on the IMSL list server in recent weeks. Geoff Nichols from IBM's Silicon Valley Lab had posted a comment stating that OSAM had less overhead, better parallelism, and Sequential Buffering. It also allowed datasets to be up to 8GB instead of 4GB limit imposed by (IMS's implementation of) VSAM. This view was supported by other postings, claiming that switching to OSAM had reduced Batch Elapsed Times by 50%. Conversion from VSAM to OSAM is simple - it just takes a DBD change, and a REORG.
We then debated 'Why have VSAM?' - various explanations/opinions were expressed - VSAM was 'strategic' in the 1980s, and with the demise of ISAM there was move to 'standardise' on VSAM for all database datasets. One advantage is that VSAM supports more extents. Another is that each time you get an extent on a new volume you get a Primary and not a Secondary.
Other advantages of OSAM were offered - better space management as a result of more efficient blocksizes (not limited to 4K multiples).
Phil Courtney - Bank of America - asked if there was a 'White Paper' or similar document containing the 'collected wisdom' on this subject. If not it would be useful to have one, as the topic seems to come up periodically.
We then discussed the various issues relating to the management of multi-volume OSAM datasets, which still seems to be problematic under some circumstances.
Next we discussed setting up OSAM Sequential buffering. Only Neil Price from TNT could offer any practical experience of this. Their approach has been to identify suitable/likely candidates, and to enable the feature where they thought it would give benefits. He mentioned one shortcoming in that Allocated Buffers are not released for re-use until end of step.
To obtain information on the effectiveness of Sequential Buffering, add a //DFSSTAT DD SYSOUT= DD statement to the job.
It was agreed that a simple guide to implementing Sequential Buffering would be useful, if anyone is aware of such a document, please send details to Frank for circulation to the group.
Sorry if your contribution to this discussion was not recorded, or has been misrepresented. I was scribbling as fast as I could!
| Topic | Comments | Score |
|---|---|---|
| IBM - What's New? | Enjoyed it! | 3.9 |
| BMC/Candle - What's New | Good to see more vendors giving a talk | 3.5 |
| Tivoli Business Systems Manager | Very interesting but would take persuading to management Very good 'whistle stop run' |
3.3 |
| IMS V7 Experiences | Good to get user's views | 4.0 |
| What's New in IMS 7.1 | Very Interesting - CF Duplexing - Something discussed
recently Good but heavy subject |
3.7 |
| OSAM & VSAM Discussion | Usefulness medium to quite good | 3.4 |