1. Our Upcoming Classes in the UK, Dallas, and Seattle
2. Upcoming System z Conferences
3. Information About Single-CP LPARs
4. DB2 Recommendations for System-Managed Duplexing
5. IBM RFE Related to DFSORT Buffering
6. Updated SMF Reference Summary
7. SMF Articles in Enterprise Tech Journal
1. New Our Upcoming Classes in the UK, Dallas, and Seattle
Based on the very kind feedback we received for our 1-day SYS1.Parmlib: Enhancements and New Features in z/OS 1.3 and 2.1 class on August 3, we are going to hold this class two more times, the first of which will be in Whittlebury, England (right next door to the famous Silverstone Motor Racing circuit!). That class will be held on November 3 (just before, and in the same location as, the GSE UK conference). The class should be of interest to anyone that has, or plans to have, z/OS 1.13 and/or 2.1 (and hopefully that covers just about all of you). In keeping with our focus on helping you get the most value from your System z investment, the class covers over 110 changes to Parmlib between z/OS 1.13 and 2.1, all of which are related to new functions, or enhancements to existing functions.
For those readers that live a little more distant from Whittlebury (Los Angeles, for example!), we are planning to run the class again just before the next SHARE in Seattle on March 1, 2015. Hopefully we will see many of you at one of these two classes.
We are also holding a new 3-day class, Exploiting New Features of z/OS to Minimize Costs, starting on December 8 in Dallas, Texas at a location to be announced. Unlike the Parmlib class which focuses on making students aware of all the recent changes to Parmlib, this class will go into much more detail on a smaller number of enhancements (some of which might have nothing to do with Parmlib). We will provide information about the rationale behind the change or the new feature (i.e. how it can minimize your costs), explain how things worked previously, how they work now, how to implement the new function, and, where appropriate, demonstrate how you would use the new function.
Our other new class is a 2-day one, starting immediately after the 3-day class, on December 11, also in Dallas, Texas. This one is called z/OS Software Pricing Strategies.Given the current preoccupation with cost-cutting, we felt that this would be an opportune time for a class about managing your z/OS software costs by fully exploiting IBM’s various pricing mechanisms and software and hardware capabilities to your advantage.
We constantly hear from customers that there is a disconnect between the system programmers that manage the systems, and the finance people that negotiate the software contracts with IBM and their other vendors. Therefore, this class is specifically designed such that every participating organization would send two people – a system programmer/capacity manager, and the contract administrator. The technician will learn about the alphabet soup of pricing mechanisms that the contract administrator has to wade through on a daily basis. And the contract person will learn about what controls the technician has at his or her disposal to control the capacity consumption of the system. We will discuss the ‘LCs’ (PSLC, EWLC, AEWLC, and so on), the new Mobile Workload Pricing (MWP), value units, understanding and verifying SCRT, capping mechanisms, Single Version Charging, Rolling 4-Hour Averages, zIIPs and zAAPs and zEDC, IBM and ISV tools, configuring to minimize software costs, and much more.
If you like more information about any of these classes, or would like to register for any of the classes, please see www.watsonwalker.com/education.html.
2. Upcoming System z Conferences
The next couple of months are going to be especially busy for us. In addition to finishing and publishing our third Tuning Letter of 2014 and preparing for our new classes:
Frank is going to be presenting three sessions at Enterprise2014 (October 6-10) in Las Vegas:
- Wed, 10/8, 1 pm – The Skinny on Coupling Thin Interrupts
- Thu, 10/9, 1 pm – Long Distance IBM Sysplex Data Sharing
- Fri, 10/10, 9 am – Why Is the CPU Time for a Job so Variable?
Frank will be presenting two sessions at IBM Systems Technical University in Atibaia, Brazil on October 28 and 29:
- Parallel Sysplex Latest and Greatest
- Why is the CPU Time for a Job so Variable?
Cheryl and Frank are both going to be presenting at the GSE UK conference in Whittlebury in the UK, on November 4 and 5
- Tue, 11/4, 12 pm – Cheryl – Why Is the CPU Time for a Job so Variable?
- Wed, 11/5, 9 am – Cheryl – SMFPRMxx Parameters: Which can Help; Which can Hurt?
- Wed 11/5 10:15 am – Frank – Long Distance IBM Sysplex Data Sharing
We hope to see many of you at these events. Please come and say hello to us.
3. Information About Single-CP LPARs
Our Tuning Letter 2014 No. 2 had a user experience titled ‘LPAR Overhead vs Risk in LPARs With One CP’ that spoke about the risks involved in running a z/OS LPAR with just a single CP. It boiled down to the fact that a system with one CP is unable to take any action to interrupt a task that loops while disabled for interrupts.
Following that article we received an email from one of our readers, Rolf Bruening of SIX Group in Switzerland. Rolf pointed out that if you detect the situation, you can generate a restart interrupt by selecting the PSW Restart option on the HMC. But he reminded us that the resulting BLW004A message is a synchronous WTOR, so make sure you see it somewhere and click on the priority button if you are on the HMC console. If you reply ABEND, ACR will terminate the disabled program and normal processing continues.
You can even write a BCPii program to issue the PSW restart command. With this method there is no need to go near the HMC in order to restart the LPAR. Rolf has tested this and assures us that it works (another good reason to enable BCPii!).
4. DB2 Recommendations for System-Managed Duplexing
Another user experience in Tuning Letter 2014 No. 2 was related to the use of Internal Coupling Facilities (ICFs) and System-Managed Duplexing. We received two emails as a result of this article.
The first was from one of our readers Scott Chapman of American Electric Power. Scott raised the very valid point that System-Managed Duplexing of the DB2 Lock and SCA (Shared Communications Area) structures was only necessary in the event of a CPC failure, and such failures are few and far between these days. Scott is, of course, completely accurate in his remarks. As with most things these days, there is no 100% right or wrong answer to this – the most accurate answer is ‘it depends’. System-Managed Duplexing of the DB2 Lock and SCA structures is insurance. If your CPC never fails, you will never need it. If your CPC does fail, the resulting data sharing group-wide outage could last anywhere from a few minutes (if your Group Buffer Pools are duplexed and you immediately restart the DB2s) up to many hours (if you have programs that run for a very long time without doing any COMMITs). I have seen both situations. My personal view is that System-Managed Duplexing is a little like an umbrella – if you have one, it will never rain, if you don’t have one, it will rain, and at the worst possible time.
The other email was about whether System-Managed Duplexing of the DB2 SCA structure was needed at all, even in the case of a CPC failure. Apparently some IBMers have been recommending for years that duplexing was not required. This was news to me. I performed tests in my previous job, creating a failure that took out a DB2 SCA structure and a connected DB2. The result of that test, consistently, was that all DB2s in that data sharing group would abend. However, I thought that perhaps this had changed in a recent version of DB2, so I checked with a friend of mine – a DB2 Distinguished Engineer that specializes in high availability. He confirmed that DB2 still works the same way – if you lose a simplex Lock or SCA structure and a connected DB2, all members of the data sharing group will abend. So the ‘right’ answer remains the same as before – whether you use System-Managed Duplexing for your DB2 Lock and SCA structures in an Internal CF configuration depends on how much insurance you want, and how important availability is to you.
Both of these valuable discussions make me think that an article about CF configurations in a future Tuning Letter might be beneficial to many of our readers. Stay Tuned.
5. IBM RFE Related to DFSORT Buffering
I was recently having a look around the IBM Request for Enhancement (RFE) database – it is amazing the interesting things that can be found there. And what else would one be doing on a Saturday night?
One of the requests that I came across was one asking IBM to improve the buffering for DFSORT SORTIN and SORTOUT data sets. The submitter (as you know, RFE does not divulge the identify of whoever created the RFE) pointed out that a DFSORT OPTION COPY job could read the SORTIN file in ¼ of the time of the SORT job. The difference between the two appears to be that SORT uses fewer buffers and does not give you the ability to increase the number of buffers.
If these numbers are accurate (and you could easily test this in your own environment), increasing the number of buffers for SORTIN and SORTOUT data sets would appear to be a good idea and hopefully IBM will place this request in ACCEPTed status.
If you do lots of sorts of large files and would like to spend less time doing that, we recommend that you logon to the RFE system and vote in favor of this request. To logon, point a browser at www.ibm.com/developerworks/rfe/. You will need to use your IBM ID and password to logon. Then do a search on DFSORT. The RFE in question is near the top of the list – number 56711.
Also, for your information, there was an APAR in 2012 that addressed a similar problem for SORTOUT files – APAR PM63848 – DFSORT Not Honoring Parameters Governing Buffer Size At Release V1R12. Hopefully everyone already has this APAR applied, but just in case you don’t, we wanted to make you aware of it.
6. Updated SMF Reference Summary
We hope that you have all seen our recently-updated Cheryl Watson’s SMF Reference Summary. It is the most comprehensive single document for describing the various SMF record types from both IBM and ISVs that we are aware of. But as word spreads about this document and its value, we have been receiving more input from IBM and numerous other vendors about their SMF records. As a result, Cheryl is hoping to complete her next set of updates and have the new version of the document available in early October. If any aspect of your job involves interacting with SMF or SMF records, this document deserves a place on your desk. You can download it free from www.watsonwalker.com/references.html. If you have any comments or suggestions or additional SMF record types or subtypes that you would like us to add, please send an email to technical@watsonwalker.com.
Cheryl Watson’s SMF Reference Summary is not copyrighted, but please leave the footer if you distribute or copy it. You are welcome to link to it, but please use
www.watsonwalker.com/references.html instead of linking directly to the PDF,
as the PDF address might change.
7. SMF Articles in Enterprise Tech Journal
Also on the topic of SMF, Cheryl and I have been working on a series of articles about SMF for Enterprise Tech Journal. This quite literally is everything you always wanted to know about SMF, maybe some things you didn’t want to know about SMF, and probably quite a few things that you didn’t know that you didn’t know about SMF. The first article, titled ‘SMF: An Important Component of z/OS’ appeared in the August/September Enterprise Tech Journal. It covered:
- SMF Introduction
- Use of SMF Data
- Major Record Types
- SMF Flow
- SMF VSAM Data Sets
- SMF Logger
- SMFPRMxx Parmlib Member
- And a list of SMF Resources
Between these articles and the updated SMF Reference, you will be able to wow your friends, exhilarate your family, and impress employers with your encyclopedic knowledge of SMF. We hope you enjoy them and find them a valuable addition to your system programmer’s library.
Stay Tuned!