Minutes of the GSE UK IMS Working Group Meeting - IBM Southbank, 12^th June 1998

Chairman: Frank Fleming (Frank.Fleming @ barclays.co.uk) - 01565-613581

Secretary: Peter Armstrong (peter_armstrong @ bmc.com) - 01276-419835

Administration

• Thanks
  • Welcome to all. Frank then went through the agenda. Dougie will review accounts later.

• Minutes
  • Those who hadn't received to contact Andrew Jack to check name and address. Want to move to electronic distribution - hence letter re email address. Please reply asap.

• Matters arising
  • Frank has to pick on non-members (or lapsed members). GSE are clamping down. Frank named the names, but I couldn't type fast enough.

• New Members / attendees
  • None.

• Correspondence
  • Frank and Peter had copies of the following for people to look at:
    • Proceedings from last year's National Conference. Call for papers for this year - ignore expiry date, can still get in.
    • GSE 1^st Technical Seminar in partnership with GSE' Spanish region in Madrid from 22-23^rd September.
    • GSE 3^rd Technical Symposium (what's the difference between a symposium and a seminar?) in Cannes, 9-10^th November.
    • GSE Newsletter
    • GSE Journal

  • GSE UK region board meeting - one item or this working group. For next year is there any interest in joint meetings with e.g. CICS or DB2 or MQ? Concern expressed that it might dilute number of attendees. CICS and MQ have joint meeting coming up.

• Apologies
  • List attached

• Other matters
  • Payment of £25 fee for this meeting to be made to Dougie please as Ian is away. Has no receipt book, so pass name and address details to Dougie and a receipt will be forwarded.

Accounts - Dougie Lawson (IBM UK)

• £2444.62 in account at end of May - part being used to fund Suzie. Amount is staying pretty static. Idea is that we set fees to cover costs, which we are doing pretty well. See Dougie's web site for details of accounts and other GSE related items like the minutes and the agenda - http://web.ukonline.co.uk/dougie.lawson/gse

IBM What's New - Alan Cooper, Pete Sadler, Dougie Lawson (IBM UK)

• Introduced our IBM guests - welcome, especially to folks from across the pond.

• Education - corrections to Alan's foils are CM50U Block Level Data Sharing is 9-11^th November. CM61U Shared Queues will be 12-13^th November. Contact Alan or Pete if you have any special requirements.

• IMS/DB2 Technical Conference in Munich next week. IMS Technical Conference in the US is in Washington in October. Pete had details of agenda for next week.

• Announcements - DIAL IBM is being withdrawn and being replaced by Public IBM iSource service. Announcement summaries (weekly update) are available via "New from IBM" at www.uk.ibm.com/nfi/home.html. Announcement letters are at www.ibmlink.ibm.com/isource/emea , where you can also set up a profile, which means you will get announcements automatically via email for selected product groups.

• New product - DBRC security tool - planned availability of August 7^th 1998. Details scant, but basically uses RACF to limit access to RECONs. Flyer claims that DBs must be registered as from V6 - we don't believe this to be true!!!!

• Flyer for IMS Recovery Saver - Pete gave us details. IMS recovery is aimed at media failure, but most recoveries nowadays are logical, and involve timestamp recoveries, which are not very friendly. Adding data sharing makes it very complicated (due to log merge) not to say impossible in some cases. Also aimed at IMS/DB2 combined recoveries. Works by conditioning a set of IMS logs and a copy of the RECONs to give you a recoverable point. Does not do the actual recovery - just prepares things for you. Can also backup RECONs with a RECORDSIZE greater than 32K.

• Red books - make sure you have level 1 version of V6 book - throw away level 0 book.

• Dougie also had details of some fixes.
  • PQ14154 - IMS DB Analyser V2 not working properly in Year 2000.
  • PQ10811, PQ13662, PQ14745. New exit in IMS, which gets called between DB read and program and between program and DB write. Designed for use with Year 2000 conversions - e.g. read 4-digit from DB and pass 2-digit to program or vice versa. You stipulate at the DBD level whether you want to use the exit for this DBD or not:

    Macros DBD and DBDGEN have been changed to allow a new keyword, DATXEXIT=YES/NO, to be specified to indicate whether the user exit, DFSDBUX1, should be called. DATXEXIT allows for a user exit, DFSDBUX1, to be used by an application while processing this database.

• A copy of the foils was distributed.

IMS Data Sharing Testing - Jim Jackson (Barclays Bank)

• Jim presented an overview of Barclays systems and requirements. When sysplex was announced, Barclays decided that their future strategy would be based on sysplex and sharing, to handle their capacity requirements, 24 hour availability needs, cost concerns etc. Things are coming back to the mainframe - putting everything on UNIX is no longer the desired solution.

• First system rolled out was a DB2 sharing system, which was implemented due to capacity problems. Went in well.

• TBO - The Big One is the one they wanted to test and is IMS V6 with sharing. (Why this one first? Capacity problems - major problem is overnight batch). Took 1000 cassettes of data to Poughkeepsie to try out 24 hours of processing.

• Batch runs had <5% overhead going from V5 PI to V6 IRLM with sharing, (and they had expected a much bigger overhead). Full data sharing with lots of BMPs actually ended up with a CPU saving. Did not test all the individual steps along the way, but appear to be getting a big saving in just going to V6. (Alan - users should not necessarily expect the same savings in their environment! May be very different!)

• Ongoing work with changing MSDBs to VSO DEDBs. Had problems with first runs due to locking, but discovered that a lot of the transactions were using update PROCOPTs unnecessarily - so they changed a lot of PROCOPTs to GO / added a second PCB. This is probably the reason for a lot of the savings.

• CPU overhead for online processing was 15-20%.

• Found 11 new problems - 5 severity 1's. 2 outstanding ones at end of testing, which were recreated at Wythenshawe. Also found 2 errors in RRDF. Good news is that they wouldn't have found some of these problems if they hadn't done the test at Poughkeepsie.

• Have proved that data sharing can solve overnight batch problem. Learnt a lot about tuning, Coupling Facility, PROCOPT changes, buffering, DR procedures. Passed all acceptance criteria apart from no open problems at end of test.

• Could you have done it all with just V6? No.

• How long was the testing? 10-11 weeks.

• Predominantly Fast Path (very little full function) and DB2.

Webbing your way to IMS - Suzie Wendler (IBM Dallas)

• Suzy presented all about how to get at IMS from the Web. First IMS system on the Web was Federal Express for package tracking.

• First consideration is whether you are coming in via SNA or TCP/IP, and the second one is whether you are going to use Conversational, Remote Procedure Call or Messaging and Queuing. The great news is that there are lots of ways you can get in, and the IMS application does not have to be changed.

• Suzie then reviewed the many different ways of accessing IMS from the Web, with details of their capabilities, restrictions, prerequisites etc. - very useful. Some of the sample programs can be downloaded from the IBM web site - go to the IMS home page at http://www.software.ibm.com/data/ims.

• A copy of the foils was distributed.

Problems forum, hints and tips, contacts for lunch

• Requirements - none.

• Hints and tips
  • Alan Cooper - IBM

    Alan told us about a large IMS user, who recently had a performance problem. At first it was thought to be a h/w (microcode) problem, but was eventually found to be an IMS "implementation" problem. CPU was proportional to transaction rate up to ca. 100 trans/sec, then the CPU went through the roof. 50 dependent regions all using Pseudo-WFI and 20% of transactions account for 80% of load. At low volumes, PWFI worked well with the top 20% getting 20-40 transactions per scheduling. But, at peak volume the top 20% only got 2-4 transactions per scheduling. The cause of the problem was that every transaction was defined with PARLIM=0. Moral - do not use PARLIM=0 for normal transactions.

  • Name: Alan Lovett - EDS - 01253-688311

    Problem: Has anyone used MVS 'RESMGR' to add DFSMRCL0 to SVC13's list of termination modules? I tried it and got S0C4 at most step ends, especially successful ones.

    Contacts/Comments: Dougie - no, can't do it - fixed in next IMS release.

LUNCH

IMS and the little Red Books - Geoff Nicholls (IBM ITSO)

• Geoff's job is to produce the IMS red books. ITSO brings in staff from round the world (from IBM, business partners and from customers) to run residencies, produce red books (and redpieces) and workshops. Redpieces are bits of books, that they put up on the web at http://www.redbooks.ibm.com, and you will also find the existing books there as well. Books are also available on CD (IBM are looking at going to Adobe Acrobat PDF files in future rather than BookManager - Peter "so are BMC" ).

• You can sign up via the Internet so that IBM will notify you if any residencies appear in your area of expertise.

IMS and the WWW (Part 2) - Suzie Wendler (IBM)

• Suzie continued (socket to me). This part also includes two comparison foils showing all the combinations with the API used, the software required, the characteristics and some basic comparisons - excellent summary. Also look at http://www.ibm.com/ims, and the web page detailed in part 1. You will find the latest information, documentation and IMS connectors - sample applications showing you how to access IMS, e.g. sample Java application (is an applet a program that fell on Newton's head?), html generation from MFS, sample ITOC (IMS TCP/IP OTMA Adapter) etc. Lots of good stuff to give you a jump start / ideas of how to code / bits to clone etc. IBM Connectors home page is http://www.ics.raleigh.ibm.com/ibmconnectors. Suzie also covered Objects and security issues.

Recovery to ANY Point in Time - Peter Armstrong (BMC)

• Peter started by explaining the architecture of the BMC RECOVERY PLUS product, which has the capability of rejecting parts of logs that are not required and sorting any log records which have not been merged (change accumulated). He then went on to explain what is happening in the IMS recovery world, which as Pete pointed out above is all about logical, application and disaster recovery nowadays, with the added complications of multiple DBMSs and n-way sharing.

• Peter reviewed the options, which BMC have already provided (recovery to /DBR NOFEOV, recovery to quiet points, creation of system wide recovery points, co-ordinated recovery etc.). This has now been extended with the ability to recover a group of IMS databases to ANY point in time, with full DBRC support of course. The product will recover you to e.g. the middle of a log and drops the inflight transactions, so that the DBs are consistent. It supports online, batch, sharing etc.

• He showed various scenarios like incorrect BMP (already happened at one customer within two weeks of announcing the product - this was the first live "disaster recovery" Peter experienced with the product), disaster recovery, rogue transaction etc. plus some scenarios which simply cannot be recovered with the standard IMS utilities.

• He also reviewed the RECOVERY MANAGER products, which enable you to automatically create recovery points, synchronise log switches, co-ordinate recovery across multiple DBMSs, check recovery assets, simulate recovery etc. These products allow you to recover any combination of IMSs and DB2s to a common point-in-time, and he showed some examples of how customers were using the products in their shops.

• A Copy of the foils was distributed.

AOB

• Duncan Mead (of Liverpool and Victoria Insurance - 01202-292333 x.2024) is beta'ing a new DBCTL monitor - contact him if you want details.

• Please volunteer to present (or volunteer your colleagues).