IMS Performance and Capacity Planning
By
Ron Steele
L. B. Software - an IBM Business Partner
Raleigh NC
919-523-0305
June 2005
Get fast and useful answers easily!
So, you don't use the IMS Monitor Reports? You're missing out on
a wealth of information, it doesn't have to take a long process to get it, and the IMS
Monitor is free (included with the IMS product).
Under the first bullet below is described how to use a short paper on "Things To Check In An IMS Monitor Report". This paper shows how to check some key IMS performance statistics in only a few minutes. Following this topic is a description of more in-depth activities you can use to analyze performance and capacity.
Free Analysis
If you are in the USA, I would like to offer you a free analysis of your IMS System using the information contained in the first bullet below, and I will report key findings back to you, with any recommendations. Just send me an email requesting the Data Collection Instructions, and I'll be happy to send them to you.
Topics covered in this article include:
The IMS Monitor Reports
I first encountered the IMS Monitor and its Reports around 1974. As an IBM Systems Engineer, I had been selling, installing, teaching, and designing for IMS for 3 years by 1974, and had encountered some problems along the way with performance. I had been to every course and seminar I could find, but had not encountered much information or help for designing for performance or tuning for IMS.
Along about 1974, an IBM FDP (Field Developed Program) became
available which was the first version of what I will now call the IMS DCMONITOR. As I was
having some problems at some of my IMS customer sites, I decided to try the DCMONITOR, and
that was one of the best moves I ever made in my 34 years of working with IMS. I
discovered very early that you could not only find performance problems in IMS, but you
could find design problems, bugs in application programs, and even bugs in IMS. IMS is a
fantastic product, but as for any large and highly useful system, it needs significant
care, and the IMS Monitor (renamed after some years) is one of the most valuable tools you
can ever have for IMS.
The Monitor was fee for some years, but was later included in the IMS
product, and is not extra cost software today.
The Monitor is valuable for checking out application program changes, and new application programs, during testing. You can discover things like:
Eventually, the Monitor was modified to support Batch-Only (DLIBATCH
and DBBBATCH) programs. The Monitor always had supported BMPs (Batch Message Programs,
both message-driven and non-message-driven. When DBCTL became available for such
environments as CICS connections, the Monitor was enhanced to support that environment
too. Also, MQSeries calls have been added to the Monitor, as well as DB2 calls and
information, and other support such as Fast Path applications.
The Monitor measures waits and the information about waits, i. e.,
what IMS is waiting on. You care about waits because the longer IMS waits on stuff, the
longer the user's response time may be. More specifically, it counts the number of waits
and measures the amount of time for these waits. As Monitor experts will quickly say, the
Monitor does more than just that, but that was the primary purpose of the Monitor, and to
me, that is the most important thing it does. For example, how many times did IMS wait for
i/o for ddname ABC and what was the mean wait time? If this time is greater than say 20
ms, then maybe the DASD tuning person may want to investigate that ddname to see if some
tuning is needed for the volume it is on, the path it is on, etc.
So, if you know what IMS is waiting on, you can investigate, and take recommended actions to minimize the wait time, the number of waits, or eliminate the waits. IMS can have to wait on many things: Format i/os, database i/os, message queue i/os, locks, etc.
Why did I write my short IMS Monitor Paper?
In 1995, I joined the IMS Advanced Technical Support Team in the IBM
Dallas Systems Center. We had several new members on the team in anticipation of the
growing demand in IBM to support the new IMS Parallel Sysplex Data Sharing enhancements
just coming out in IMS V5. The Old Timers on the team were gracious to hold training
sessions for us New Timers on the team, and one of the things they wanted was for us to
all be using the IMS Monitor in similar ways so that we didn't confuse customers with
different approaches. At that time, I documented the approach taught, and published it in
the web for others to have public access to. I left the team in 1998 to work on some
Tivoli consulting for a few years. I discovered that in the early 2000s, my IMS Monitor
paper was still there on the web. What was more, I continued to get questions about the
paper, and found that it was still widely used, even though I was not able to update it
for later versions of IMS.
Here it is mid-2005, and it is still significantly used. It has been
a good reminder to me of some IMS stuff as I returned to IMS fulltime in 2004 as an
associate of L. B. Software. As I retired from IBM in 2001, I thank IBM for leaving the
paper on the web as many IMSers still use it. I have some concern that it has not been
updated, but it is very valuable to many IMSers as it is. I hope to get the
time someday to bring the paper up to date for IMS V9.
It is a good quick check even today for a health check of IMS
performance. You can go through a Monitor Report (even a several hundred page report) in a
few minutes, or less than an hour, and learn almost any significant performance problems
by following the sequence stated in the paper. Click here to retrieve
the paper. As I am no longer at IBM, please use the email id above to contact me, instead
of the email id in the paper.
I would like to make one suggestion when using the paper: "Check
the General Program Statistics" after checking the "Mean I/O IWAIT Time
for whole system". Then come back and check the Buffer Pools. This
will allow you to see if there are response time problems sooner in your analysis. If you
are satisfied with the response times, then you could stop your analysis here.
Using this quick check is not as good as using the Redbook described in the next topic, but the redbook is 510 pages and my paper is about 15 pages. So, how much time do you have? I recommend using the redbook if you have time. If you don't have time to use the whole redbook, you can use the redbook as a reference document, and to see what has changed up through IMS V7.
IMS Performance & Tuning Redbook
The first IMS Tuning Guide was published in about 1977. I recall trying to read my first copy of it, and watch Rocky I at the same time, on a flight back to Washington from Palo Alto (One of the homes of IMS Development over the years). I guess I'll always associate tuning IMS with Rocky I...among other things. The first guide was great because it was the first time I saw ROTs (Rules of Thumb) for IMS performance. For example, mean i/o time should not exceed 30 ms, or whatever it was back then. There were other valuable ROTs given too. It also explained the Monitor Reports excellently, giving a different view of IMS performance than the Monitor manual. It contained flowcharts of how to go through the Monitor Reports and did an excellent job of providing a structured approach to analyzing IMS performance. As I recall, it even contained information on finding information in RMF Reports that affected IMS performance.
This guide has been updated many times over the years, and the last
update I am aware of is this redbook, IMS Version 7 Performance Monitoring and
Tuning Update: Click
here to retrieve this redbook. I highly recommend this book as a tutorial
on IMS performance.
Now, you say, what about IMS V8 and V9 coverage in this redbook? I
have made a request to IBM to update this redbook for IMS V8 and V9 as soon as possible. I
will keep you up to date on the update for this redbook.
IMS Performance Analysis Tool
The following is reprinted from the IBM Tools website: http://www-306.ibm.com/software/data/db2imstools/imstools/imspa.html
. I do not have experience with this tool, but for as much completeness as possible, I
want to include all IMS performance information I am aware of.
IMS
Performance Analyzer for z/OS, V3.3 provides comprehensive performance analysis and tuning
assistance for IMS TM and DB. IMS Performance Analyzer for z/OS processes IMS Log,
Monitor, and now IMS Connect event data to provide comprehensive reports for use by IMS
specialists to tune their IMS systems and managers to verify service levels and predict
trends.
The following table shows the versions
since November 1997:
Ver/Rel |
OS/Platform |
PID |
GA |
EOM |
EOS |
Replacement
Product |
3.3.0 |
z/OS |
5655-E15 |
TBD |
TBD |
Not
Established |
|
3.2.0 |
z/OS |
5655-E15 |
IMS
Performance Analyzer v3.3 |
|||
3.1.0 |
z/OS |
5655-E15 |
IMS
Performance Analyzer v3.2 |
|||
2.1.0 |
z/OS |
5697-F02 |
IMS
Performance Analyzer v3.1 |
|||
1.1.0 |
z/OS |
5697-B89 |
||||