1. Cheryl Watson’s TUNING Letter 2002, No. 1 Summary
3. CICS Response Time Spikes
4. CD-ROM Searching
1 . Cheryl Watson’s TUNING Letter 2002, No. 1 Summary
The fifty-two page 2002, No. 1 TUNING Letter was emailed to electronic subscribers on April 3. The issue will be mailed to print subscribers on or before April 12. Here are a few of the topics we covered, taken from our “Management Issues.” You can purchase a printed copy of the current TUNING Letter for $85.
In February, IBM announced their latest z/OS machine, the z800 processor. With its capacity range of 80 to 635 MIPS and its ability to run the latest releases of z/OS, it provides an excellent option for small to medium sized installations. And for larger installations, the pricing of the z800s make them very attractive for running your e-business work without increasing the cost of your legacy software holdings. We discuss two new pricing options which provide software at a 5-35% reduction and even a 90% reduction (for e-business use)!
In February, IBM changed their LSPRs (benchmarks for hardware capacity) to include the z800s. At the same time, they updated the LSPRs for some older machines (Multiprise and G2/G3/G4s). Because both the workloads and the software releases changed for the benchmarks of these older machines, the apparent MIPS ratings of the older machines was increased. This is particularly noticeable if you’re trying to compare HDS or Amdahl machines to the IBM machines. See our article on page 42 which explains what’s happening.
The performance of the Hierarchical File System (HFS) used by Unix System Services in OS/390 and z/OS has not been particularly impressive. IBM has now announced a new zSeries File System, zFS, which complements, but doesn’t replace HFS. It is designed to provide much better performance (30-50% better) than HFS for certain types of files. In this issue, Clark Kidd of Allen Systems Group provides an introduction to HFS, and instructions on how to install it. See page 31 for this focus article.
After Goal Mode
Most installations will have migrated to WLM goal mode by now. If you haven’t done so already, you should be aware that the latest z/OS release (1.3) no longer supports compatibility mode. At SHARE, it looked like less than 20% of the sites were still running compat mode. It’s time to convert! And if you are running in goal mode, there are plenty of new facilities that will make management of your system much easier. Our second focus article in this issue describes which features have been well received and which have barely been used. See page 15.
Elsewhere in This Issue
Much of this issue is taken up with news from the recent SHARE conference in Nashville. See “Cheryl’s Hot Flashes #7” on page 8 for my latest tips, my SHARE Trip Report on page 26 for information heard at SHARE, and Bob Archambeault‘s CICS SHARE Trip Report on page 45. Installations going to 64-bit may get some surprises (pages 4 and 44).
Here are some updates on prior items:
a. My SHARE presentation, “Cheryl’s Hot Flashes #7,” had an incorrect APAR number on page 35. APAR OW46377 should be changed to OW46477. Thanks to Tom Conley of Pinnacle Consulting Group for pointing it out.
b. The updated March 2002 CPU Chart was sent to electronic subscribers on March 21. The print issue was mailed on April 5. (Sorry for the delay, but the printer goofed!) There is still a correction needed for the electronic edition, but since it affects only a few people, we decided not to send out one more 600K email. The LSPR release for the 9672-Rx1 series should be R1 instead of R4. If you’d like an updated CPU Chart with that correction, please send an email to <email@example.com> including your company and address. Also, the November 2000 CPU Chart had an incorrect release specified for the G1, G2, G3, and G4 machines. Those should all be R1 instead of R4.
c. Cheryl’s List #60 (and also in our TUNING Letter 2002, No. 1, page 49) made a reference to APAR OW45117 by saying that it was good to apply for CICS and IMS regions that are managed with transaction goals. Steve Rowe of DST Systems pointed out that he was told by IBM that it really applied to any address space that is blocked for CPU. Therefore, it’s a good APAR to apply in all environments.
d. In our TUNING Letter 2002, No. 1 (page 26), I mentioned APAROW49983. I had indicated that it was PE with a correction by OW53001. It is not PE, but corrects another APAR. OW53001 is the rolldown of OW49983 to the OS/390 R6 release. I also mistyped two APARs on page 27. PAV APAR OW59243 should be OW50243, and RMF APAR OW49636 should be OW49536. Page 47 also had CICS APAR PQ40571, which should really be PQ45071. Many thanks to Jerry Urbaniak of Acxiom CDC for pointing these out.
3. CICS Response Time Spikes
In the last Cheryl’s List (#64), I described a situation where a site was experiencing spikes of long CICS response times whenever SMF wrote records on the synchronized interval. Usually the problem with this is that you have DDCONS(YES) coded in SMFPRMxx. But this site didn’t have that specified. The problem has been resolved, and we thought you might like to know what caused it.
It seems that someone changed the weights of the LPARs incorrectly. The CICS LPAR was effectively weighted at 30% of the machine, when it used to be 50% of the machine. During the SMF interval, SRBs are scheduled to all address spaces in order to collect data, and the CPU runs at 100% on all LPARs at the same time. During this collection period, the CICS LPAR was capped to 30% instead of 50% of the machine. Adding to the problem, all ten CPs were assigned to the LPAR, so that CICS would tend to get only 3% of a processor at a time. Because the CPU spike only occurred for seconds, the RMF reports didn’t show the spikes. This is just one more reason to only assign as many LPs (logical processors) as you absolutely need.
4. CD-ROM Searching
While searching our CD-ROM for ’64-bit’, I noticed that it wasn’t finding some articles that I knew I had written. It turns out that when using the Adobe global search for multiple documents (i.e. global search) on the CD-ROM for anything with hyphens, you need to remove the hyphens. So a search for ’64bit’ gives you all thirty references to ’64-bit’, but a search for ’64-bit’ only gives you 12 references. When you’re in a single document, however, you must include the hyphen or you won’t find any matches. Adobe has been notified and they are addressing the issue.