1. Cheryl Watson’s Tuning Letter 2008 No. 3
2. Updates to TUNING Letter & CPU Chart
3. SHARE in San Jose
4. 2009 Newsletter Rates
5. Humor from our Son-in-Law, the Engineer
1. Cheryl Watson’s Tuning Letter 2008 No. 3
The fifty-two page 2008 No. 3 Tuning Letter was emailed to subscribers on July 24. You may visit our Web site at http://www.watsonwalker.com to obtain subscription information. The following is a summary of just some of the contents of this latest Tuning Letter:
IBM Announcements
The most important management news in this issue is the announcement that IBM recently acquired Platform Solutions, Inc. (PSI). PSI has developed an Intel-based machine that can run z/OS, Linux, and Windows. This acquisition may not impact most of our readers directly, but we’re hoping that it will help the software developers and their sales forces by allowing z/OS to run on a less costly PC or laptop platform.
HiperDisptach
The newest machine in the zSeries, the z10, provides a hardware/software solution called HiperDispatch. HiperDispatch reduces the overhead of the MP (multi-processing) effect. The new design can provide up to a 10% improvement in performance, especially for the very large installations. I think that it’s a great design, and it takes minimal effort to implement.
SMF Update
This issue provides a parameter-by-parameter review of SMF options. Because SMF (System Management Facility) is the mechanism that provides almost all of our computer information, it’s important to ensure that it’s healthy and running at optimum efficiency. Reviewing the parameters periodically is one way to ensure that. This issue continues with our SMF Update series of articles. One important note is that in the last issue, I said that the SMF Logger wasn’t ready for production because it didn’t have a CLEAR facility. Now that a tool is available to solve that, I can recommend that z/OS 1.9 sites investigate SMF Logger.
z/OS 101
The second installment of our new series, z/OS 101, covers some basics of SMF for those people fairly new to z/OS (page 35).
Elsewhere in This Issue
Our User Experiences on page 39 highlights some suggestions when going to z/OS 1.9. New Function APARs shows that IBM is now moving even more services (this time it’s XML) to zIIP processors. Expect to see more along this line. Some users have had problems with discretionary work on z/OS 1.8 swapping out and taking a long time to swap back in. On page 43, we include an APAR to correct that. In addition, we’ve included a long list of HIPER APARs that should be of interest. Many of these APARs describe problems where data is lost (early truncation, etc.). If you see an ABEND or system failure, you can easily search the IBM database for a solution. But lost data may not be discovered until it’s too late to recover (if it’s discovered at all). That’s why I think it’s especially important to keep HIPER DATALOSS APARs as a high priority.
2. Updates to Tuning Letter & CPU Chart
Correction to HiperDispatch Article
In the last issue (2008 No. 3), while discussing HiperDispatch on page 32, I said that IRD changes the SRM constant when configuring CPs offline and online. That is not true, as Jim Horne of Lowe’s Companies pointed out to me. Jim has run many tests on various configurations which prove his point that the constants are not changed. Jim published his findings of IRD, zIIPs, zAAPs, and RMF/CMF measurements in an excellent CMG paper called The Myth of MSU or, How Big is the Bucket? If you’re a CMG member, you can obtain the paper at http://www.cmg.org/membersonly/2007/papers/7171.pdf. Jim’s SHARE presentation on the same subject is available to SHARE members by going to session 2519 in the Orlando proceedings. The
link is http://ew.share.org/proceedingmod/abstract.cfm?abstract_id=16864.
Feedback on Last Tuning Letter
The articles about HiperDispatch, SMF Logger, and the SMF parameters generated a LOT of feedback – all of it appreciated. Because of their circumstances or experience, some people do not share my enthusiasm for the SMF Logger or HiperDispatch, and other people have reasons why they would not use my SMF recommendations. I welcome that feedback, and plan to include our readers’ comments in our next issue. So, if you have something to add, please send it along.
Request for Help
Two of our readers have asked questions that I can’t answer, but thought that other readers might be able to help.
One person would like to talk to anyone who is using a DVIPA (Dynamic Virtual IP Addressing) Sysplex Distributor for DB2 in their installation. Does it distribute the work correctly for you? (And I and others would be interested in any of your conclusions too!)
Another reader would like to know the best way to set goals for the ADSM Server – discretionary or high, as in SYSSTC. I found one forum that discussed this, and answers were all over the board. Would you share any experiences that might be helpful?
HiperDispatach=NO on z10
A reader just converted from a z9 to a z10, left the HiperDispatch default of ‘No’, and continued to use IRD with VARYCPU and CPU management enabled. They have found that the behavior of IRD on the z10 machines is not as they expected. It seems that the logical processors are being managed differently. After an IPL, the number of engines online is drastically reduced and rarely come back online.
Have you heard of any problems with IRD after migrating to z10?
3. SHARE in San Jose
I’ll be heading to San Jose on Saturday for the upcoming SHARE conference (August 10-15), and I hope to see many of you there. For information about the conference, go to www.share.org. If you can’t attend the conference, you can still view most of the presentations by clicking on ‘Proceedings’ on the left, then selecting ‘San Jose‘ in the drop-down. The presentations are put up early by some speakers, but should all be available by August 19th, which is the cut-off date. Several speakers, like me, tend to wait until after the presentation in case there are changes to be made.
I’ll be presenting at two sessions, and will be part of a panel at another:
Session 3117 (with Linda Mooney) – Can You Teach a New Capacity and Performance Specialist Old Tricks? Hindsight and Insight, Wednesday at 1:30 pm
Session 2509 – Cheryl’s Hot Flashes #20, Friday at 9:30 am
Session 2549 – Meet the Experts: WLM, Performance and Capacity Planning Topics, Thursday at 4:30 pm
If you’re new to z/OS, be sure to check out the zNextGen project and look for their favorite sessions. If you’re an old hand at z/OS, and are coming to SHARE, be prepared for a huge number of sessions on z/OS 1.10. Because it was only announced on Tuesday of the February SHARE, most sessions there had only a small amount of z/OS 1.10 information. At this SHARE, most of the sessions will have the latest information. Obviously, I’ll be including many of these topics in future newsletters.
4. 2009 Newsletter Rates
Many of you have asked for the 2009 Tuning Letter and CPU Chart rates. Starting in January of 2009, the single-site subscription will increase by $85, from $865 to $950. This represents an increase of 5% per year since our last rate increase two years ago. If you’ve considered subscribing to the newsletter, be sure to do it before the end of the year.
We are also making one major change to our rate structure: We are no longer selling individual newsletter issues or single copies of our CPU Charts. Please check our Web site for a complete listing of the 2009 subscription rates and for the new CPU Chart pricing options.
5. Humor from our Son-in-Law, the Engineer
There are 10 kinds of people in this world. Those who understand binary, and those who don’t.
Stay Tuned!