In this issue, I’ll cover the following:
1. Amdahl New Processors
2. HDS Skyline Processor Groups
3. London class canceled
4. Cheryl Watson’s TUNING Letter, 1997, No. 6 Summary
5. The next TUNING Letters
6. New on our Web Site
7. More on Catalog CPU Overhead
1. Amdahl New Processors
On February 9th, Amdahl announced the general availability of their CMOS Millennium model GS700 series (based on a 75-MIPS uni-processor) beginning February 20. This announcement came with a surprise since they also announced the April availability of a new set of processors based on an 80-MIPS uni-processor. All of the installed 75-MIPS processors will be upgraded to the faster speed CPU. And this is for a CMOS CPU! This is really exciting news. It means that CMOS is truly a replacement for the traditional bipolar machines.
The smallest of the new Amdahl series is the GS715 at 80 MIPS, software group 40, with 14 MSUs. The largest model is a 12-way GS7Z5 estimated at 686 MIPS and 119 MSUs. Amdahl indicates that this makes their CMOS competitive up through the HDS Skyline 527. That’s pretty impressive! In the same announcement, Amdahl also introduced two down-graded processors, the GS732 and GS782, to fill in their processor group range.
The new Millennium models are:
We’ve updated our CPU Chart with the MIPS for these new models, and a new chart will be sent to all subscribers by the end of April. Subscribers to Cheryl Watson’s TUNING Letter may ask to receive a fax or email containing MIPS for the new models by sending an email to doni@watsonwalker.com (be sure to include your company name and address).
Amdahl’s announcement (along with the stories of the yet-unannounced IBM G5 series (9672-Rx6) that is supposed to be based on a 90-MIPS uni-processor with an expected availability date of 4Q98) seems to ensure that even the largest installations will be implementing CMOS processors at some point in the (near) future.
2. HDS Skyline Processor Groups
When we published our last CPU Chart, we did not have the SU/Sec or the version code on the latest HDS Skylines. The MIPS listed in the CPU chart are accurate.
The HDS Skyline 400 series:
We’ve updated our CPU Chart with the data for these models, and it will be sent to all subscribers by the end of April.
3. London class canceled
Due to time constraints, we are unable to hold our recently announced Advanced class in London September 14-18, 1998. Perhaps next year will be less hectic. We apologize to any students who intended to sign up for London.
4. Cheryl Watson’s TUNING Letter, 1997, No. 6 Summary
This issue was mailed to subscribers on 24 February, 1998. The following is an extract from the Management Summary to show you the scope of the issue.
Tuning Parallel Sysplex
Parallel sysplex can provide continuous availability for critical applications and can provide recovery scenarios that result in the minimum exposure to the customer. However, implementation of parallel sysplex is a large investment for any installation. Managing and tuning the parallel sysplex resources, then, becomes important to provide the maximum benefit for the cost and to ensure that performance is not degraded.
Our focus this issue provides an in-depth discussion of how to tune the parallel sysplex resources such as the coupling facility and how to provide the best configuration in an LPAR environment. The article is written by Joan Kelley, one of the most knowledgeable people we’ve met on tuning parallel sysplex. Joan works with the Parallel Sysplex Test and Evaluation group in Poughkeepsie, and has helped many customers tune their installations. We think that implementation of the techniques provided in Joan’s article can save you thousands of dollars in better managed resources.
A note on parallel sysplex has been added on page 6, from Mike Cox, Senior Consulting IT Specialist of IBM’s Washington Systems Center, one of the most enthusiastic supporters of parallel sysplex. Mike has helped several dozen customer installations with parallel sysplex and feels that most of them lack three things needed for successful implementation: 1) understanding of XCF transport classes, 2) a Sysplex Failure management (SFM) policy and procedures in place for structure rebuilds, and 3) an Automatic Restart Management policy or multi-system automation procedures for restarts. Are you in the same situation?
Dynamic LPA
Dynamic LPA (Link Pack Area) is a new feature that became available with OS/390 R4. It is one of the major steps toward allowing an image to go for months between IPLs, while still permitting maintenance updates to LPA. The new facility isn’t well advertised, so we’ve asked Bob Shannon from Programart to provide an in-depth discussion of this facility on page 25. If you are trying to find ways to delay IPLs, this is a major enhancement from IBM. (Note that there have been other solutions from software vendors to provide a similar facility for several years, but this is the first IBM solution.)
Catalog Overhead
For several issues we have been updating our readers about APARs related to high CPU overhead in the catalog address space (CAS). New fixes appear almost every other week. We’ve provide at least three new things to try on page 5 and 36.
Software Pricing
In previous issues, we’ve brought up the caution that OS/390 V2R4 may result in higher prices due to the increase for the version change. We’re now hearing from several sites who say that their total software charges have actually gone down when they moved from V1 to V2. It’s still something you should check before the change.
SAS Institute’s IT Service Vision
For installations who are trying to resolve the issues of managing and analyzing huge amounts of data, SAS Institute’s IT Service Vision may be just what you need. It’s the master of data reduction for all types of data warehousing applications including your performance and capacity data bases. Our product review is on page 33.
Licensed Manuals Available
If you are part of a consulting group or work for a software or hardware vendor, you may have been inconvenienced because of the lack of access to many of the IBM licensed manuals, such as the data areas manuals and diagnosis guides. IBM is now changing many of these previously licensed manuals to unlicensed, and many are even available from the Web. Page 10 has the Web address that describes the manuals now available.
Elsewhere In This Issue
In S/390 News starting on page 5, we describe the following: a new OS/390 R3 command to display LLA, a warning to OS/390 R3 installations to avoid multiple LNKLST extents, four APARs to reduce paging and swapping in goal mode, an APAR to reduce problems with CICS sockets in TCP/IP, a new facility for IOS called I/O Timing Message-Only service to allow timing of DASD, an APAR about severe slow-downs using ERASE ON SCRATCH in DFSMS on ESCON channels, a pointer to WSC flash 9748 (OS/390 R4 Firewall Technologies Support), a pointer to WSC Flash 9748 (Using a Coupling Facility for the JES2 Checkpoint), and pointers to a Web site for redbooks (abstracts, full online published manuals, and redbooks in progress).
Our schedule on page 38 and our class schedule on page 35 show our 1998 classes. One class is taught in Toronto this year and the rest are held in Sarasota. If you haven’t received our class brochure, please call.
A simple parameter in IEASYSxx, RSU (Reconfigurable Storage Unit), is causing some sites unnecessary CPU overhead when they upgrade a machine. The symptoms are difficult to see, and you may have been running in a degraded mode for a couple of years without knowing why. Our update on our Parmlib Series on page 31 shows you how to calculate what the value for this important parameter should be.
5. The next TUNING Letters
The next issue of the TUNING Letter, 1998 Number 1, is expected to be mailed in the next two weeks. The following issue, 1998 Number 2, will be sent out within a couple of weeks of that one. We apologize for the delay, and appreciate your patience. This is the year we WILL catch up!
6. New on our Web Site
We’ve added several new items to our Web site. Now you can subscribe to the TUNING Letter online. In order to help you select back issues, we’ve expanded the tables of contents of each issue. You can also register for classes online. Other improvements are in the works.
A revision of my paper on “Why Your Capacity May Not Match Your Vendor’s Estimate” can also be found on our Web site at http://www.watsonwalker.com/lspr.html. This paper was first published in the 1997, No. 4 issue of Cheryl Watson’s TUNING Letter, then presented at CMG ’97 and later as a four-part series in Technical Support magazine.
7. More on Catalog CPU Overhead
Rene Belanger, of Chrysler Corporation, told us about one more APAR that seems to have provided relief for this ongoing problem. His shop was running under MVS 5.2.2 and installed VTAM 4.4 at PUT9710. The CPU time in CAS increased by 2 to 3 times, batch jobs ran longer, and TSO response time was degraded. This occurred in images that ran DFHSM data migration. A circumvention was to issue a restart to the catalog, but the PTF for APAR OW30422 (with child OW31322) also fixed the problem. This is a VTAM V4 APAR.
Joan Mulcahy, of M&T Bank, told us of another possible fix for high CAS CPU time. They noticed the increase after migrating to SP 5.2.2 and DFSMS/MVS 1.3. They received a note from Sterling Software that a performance PTF, SS02917, was available that would reduce the number of SC 26 calls (CATALOG, INDEX, LOCATE macro) and improve performance for the SAMS storage management products. Prior to applying the PTF, they were using 37 to 42 minutes of CAS CPU time between 8am and 5pm. After the PTF, the usage dropped to 30 to 36 minutes. That’s a nice savings!
That’s all for now. Stay tuned!