1. COBOL V5 and LLA and PDSE, Oh My
2. SHARE, Here We Come
3. Classes in New York City
1. COBOL V5 and LLA and PDSE, Oh My
You might recall that in Cheryl’s List 181 we published a little information about the performance experiences of one of our clients when they moved to COBOL V5. That article generated quite a flurry of emails from other readers and from three IBMers in particular – Peter Relson, Tom Reed, and Tom Ross. I would like to take this opportunity to thank them for their help, their openness, and their endless patience.
Thanks to them, I have learned more about COBOL, LLA, and PDSE than I thought possible. As my kids used to say to me – “TMI” (Too Much Information, for those of you that are my generation). But it has also been very interesting. It was a little like peeling an onion – every answer from Peter or Tom or Tom brought new questions from me.
But, first, to put your mind at rest, we believe that there is no systemic performance problem in COBOL V5. We believe that the root of the problem was that this customer had not enabled the PDSE Hiperspaces. They use LLA to manage their application load libraries, and prior to COBOL V5, the program objects were eligible to be placed in VLF. However, COBOL V5 program objects contain deferred segments, and modules containing deferred segments are currently not eligible for VLF (but look for LLA in the z/OS 2.2 preview for a planned enhancement in this area). IF they had had the PDSE Hiperspace enabled, LLA would have cached the COBOL V5 program objects in the Hiperspace and the customer might not have even noticed any performance difference. However, because they didn’t have the PDSE Hiperspaces, the modules could no longer be cached, resulting in the significant increase in I/Os to the load libraries that they observed.
So, if you have already embarked on a migration to COBOL V5, or are about to, there is no cause for concern, as long as you make sure that the PDSE Hiperspaces are enabled (use the D SMS,PDSE(1),HSPSTATS,SUMMARY command to check). The default is that they will not be enabled, so don’t take it for granted that they have been enabled in your systems. If they are not, you can enable the SMSPDSE1 Hiperspace dynamically using the SETSMS command and then restarting the SMSPDSE1 address space. The SMSPDSE address space is not restartable, so it can’t be changed dynamically. However SMSPDSE is only used for LNKLST libraries and you probably do not have your application load libraries in LNKLST, so that should not be an issue.
As I said above, in the process of looking into this question I learned a LOT of useful information about PDSE, LLA, and COBOL (and other programming languages). What started off as a one paragraph answer turned into an 18-page article that will be contained in the next Tuning Letter. Just to give you an idea of some of the questions that I had that we will be covering in that article:
- What controls which modules get cached in the PDSE Hiperspace?
- What metrics are there to understand how the PDSE Hiperspace is being used?
- What metrics/tools are there to see how VLF is being used?
- Why does COBOL V5 create modules in a format that currently can’t be placed in VLF? And is it only COBOL V5 that does this? (The short answer is NO.)
- Is there any relationship between this issue and the page-fault-mode that is described in Information APAR II14026?
- If you have a choice between caching a program in VLF or the PDSE Hiperspace, which is ‘better’?
- There is an old Redbooks tip
(http://www.redbooks.ibm.com/abstracts/tips0567.html?Open) that says that ONLY PDSEs with MSR=1 can use the PDSE Hiperspace. Is that still true (the tip is nearly 10 years old)? And how does that interact with LLA? Can LLA decide that something should go in Hiperspace regardless of the MSR value?
Also, hot off the presses – IBM just published a technote titled An Overview of Hiperspace Caching for PDSE that discusses PDSE Hiperspaces in general – if you don’t have much experience with PDSE Hiperspaces, you should definitely review that document.
While we are on this topic, we have a few questions about how you use COBOL and LLA and PDSE – if you get 2.489 minutes free, could you please complete this short 5question survey? We will feed back the results in our next Cheryl’s List.
2. SHARE, Here We Come!
As I write this Cheryl’s List, I’m just finishing off my schedule for SHARE next week.
Well, OK, that’s a bit of a lie. I’m trying to finish my schedule. But there are just TOO MANY sessions that I want to attend. Of course there are sessions on the new z13 that are a must-see. And then there are the sessions related to articles in upcoming Tuning Letters. And some sessions on mainframe hacking – I can attend these without suffering sleepless nights now that I don’t look after production systems any more. And of course there is John Eells session on z/OS 2.2 and Marna Walle’s sessions always contain tons of useful tidbits. And SHARE would not be SHARE without catching one of Bob Rogers’ sessions. And, the three gentlemen (Peter, Tom, and Tom) that helped us with our COBOL/LLA/PDSE questions will all be there and all presenting, so I need to see their sessions and thank them for their help. There is one slot where there are five sessions that I would love to see. And there is not a single slot when there isn’t at least one session that I would like to attend. In fact, there are a few sessions that are on at the same time as my own session that I would like to attend!
Speaking of my sessions, if you are going to be at SHARE, Cheryl and I would love to meet you. We will be presenting the following sessions:
- Monday 15:15-16:15 – Frank – The Skinny on Coupling Thin Interrupts. This has been significantly enhanced since the session with the same title in Pittsburgh.
- Tuesday 16:30-17:30 – Cheryl and Frank – SMFPRMxx Parameters – Which can Help; Which can Hurt
- Friday 10:00-11:00 – Cheryl and Frank – The Cheryl and Frank zRoadshow
If you are going to be at SHARE, take a few minutes now to download the SHARE app to your smartphone, or go to https://share.confex.com/share/124/webprogram/start.html and click on Create an Itinerary to create your own itinerary. And please come along to one of our sessions or stop us in the hallways to say hello.
If you are not going to SHARE, consider signing up for SHARE Live. This lets you attend specially selected sessions in real time remotely for a reduced fee. You can also download selected sessions for up to six months after the conference. For more information, refer to http://www.share.org/seattlelive. In my humble opinion, it is no replacement for the real thing (sort of like alcohol-free Guinness or fat-free ice cream), however it is still a valuable experience.
3. Classes in New York City
Speaking of valuable experiences, we just wanted to remind you that we have two classes coming up in New York City in April. The first one is a two-day class about the perennially popular topic of cost savings, starting on Monday, April 20. There is a bewildering array of things that you can do to reduce the total cost of ownership of your z/OS systems and we plan on covering most of them in our z/OS Software Pricing Strategies class. So the next time your favorite accountant complains about ‘the cost of the mainframe’, point him or her at the description of our class (at http://www.watsonwalker.com/Pricing.html) and tell them that if they are really serious about reducing the cost of running z/OS systems, then they should send you to the class.
The other class is our 3-day Exploiting New Features of z/OS to Minimize Costs class, starting on Wednesday, April 22, immediately after the software pricing class. Sadly, many sites are not exploiting many of the new features that are provided in the latest z/OS releases. Many sites operate on the basis of: install the ServerPac, make sure everything works the same way that it worked previously, return to day-to-day firefighting. However this is both a shame and is also not cost effective. Every new z/OS release brings a plethora of enhancements, and exploiting even a subset of them can save you money, improve your response times, and give you better availability. So if you don’t exploit these capabilities, you are missing all those opportunities. And remember, you are paying for all these features as part of z/OS – if you don’t use them, IBM does not give you a refund for the parts of z/OS that you don’t use.
And to gladden the heart of even the stingiest purse-string holder, we are offering a discount of 50% for the second and subsequent students that enrolls from each company. We are firm believers in the value of education, so we are doing all we can to encourage companies to send as many students as possible.
We hope we’ll see you at one of our classes. If you have any questions, please send us an email at email@example.com.