Performance problem

Posted on 2012-09-20
Last Modified: 2012-09-27
Hi Experts,

I have a Performance problem, some times of the day, the I5 is really slow, while CPU activity is only 30%.
Can someone tell me what to collect (and how) to get useful information within 1 or 2 day's

Question by:theo kouwenhoven
    LVL 13

    Expert Comment

    Hi Murph

    During the performance problem, run WRKACTJOB, WRKSYSSTS, and WRKDSKSTS. Let 5 - 10 minutes pass and refresh each of those commands.  Look for excessive disk arm busy levels and memory faulting.
    Also use WRKSYSACT to see what jobs are running.

    Can you confirm that it is the server and not the network that is slow?
    Does poor performance occur with all interactive jobs or some interactive jobs? batch? ODBC?

    You can also use performance tools to collect information.  GO PERFORM and check out the options. The default interval of 15 minutes is usually fine.
    After the collection has run during the problem time, use DSPPFRDTA and ANZPFRDTA to see what is going on. The best bet is to pick an interval with the problem and analyze/display the details.

    Hope this helps!
    LVL 16

    Author Comment

    by:theo kouwenhoven
    Hi Barry,

    I don't like refresh the screen every 5 minutes for 24 hours :-) so I need to collect ANZ data.

    GO PERFORM  will not work, because the Analyzer is not installed, we have to collect the data and do the analyzing on an other system.

    What do I have to start (what parms and what config) to get a good base for the analysis.
    LVL 34

    Accepted Solution

    Slow system with only 30% CPU utilization is generally either due to disk overutilization (WRKDSKSTS during a slowdown and look for disk units exceeding 40% busy to confirm), or memory limitations (WRKSYSSTS and look for high page faulting rates).  There are other possible causes: very high network utilization can make interactive jobs seem slow, various type of object contention, etc.

    STRPFRCOL command can be scheduled to create a performance collection each day.  Be aware that there is overhead associated with performance collection, so you may make things a bit worse - but it is necessary unless you can learn what you need to by direct observation.

    I suggest using the STANDARDP collection profile:


    This will generate a *MGTCOL object that you can send to another system for analysis with Performance Tools or iDoctor.

    Performance data collections can get very big, depending on the size of your system and amount of work in the system.  Every time you run the command above, the system will roll over performance data into a new *MGTCOL object.  On a small system, I like to roll performance data over once a day at midnight, so I have all the data for a given date in one management collection.  On a larger, busier system, you might want to roll over more often in order to keep the MGTCOL object sizes to a more usable size.

    You can also download a free 45-day trial of the excellent iDoctor tool - which use almost every day - which collects detailed performance data and provides great tools for analyzing performance problems.

    There are a long list of tools used in performance analysis and investigation, depending on the nature of the problem, but I usually start with Collection Services or idoctor Job Watcher, and then decide where to go from there.  The System Performance Management reference guide covers the basics of system performance analysis:

    You may already be collecting some performance data and sending it to IBM if you use PM/400 (GO PM400)  This service collections performance data and sends it automatically to IBM.  See page 95 of the Performance guide above for details.

    Performance Collection and PM are related, so make sure you understand the relationship and data location requirements if both are to be used at the same time.

    - Gary Patterson

    - Gary Patterson
    LVL 16

    Author Closing Comment

    by:theo kouwenhoven
    Excellent as always :-)   Thanks
    LVL 26

    Expert Comment

    You might also want to check what any job is doing when jobs become slow. "Performance" might even be involved.

    Say, for example, that one job runs at intervals and sets a series of exclusive locks on various objects that are widely used. As each lock is set and released a moment or two later, other jobs pause and wait. There might be little disk or memory usage that is out of the ordinary, but lots of jobs can queue up to get their locks.

    A specific case could be the database files in QSPL. Say you have some security/compliance product that periodically collects ownership/authority info for objects around the system, and someone configured it to include objects in QSPL. (I've seen this case and similar ones.) A lot of spooled files on the system will result in a lot of members in the QSPL database files. And because authority is technically held at the member level (even though applied at the *FILE level), the member index can be locked until the APIs finish the work with the members. Jobs that want to spool joblogs or open report (printer) files will simply wait. System resources will look otherwise perfectly normal.

    Similar conflicts can happen in other libraries where files have large numbers of members or in other cases. The point is simply that knowing what one or more of the unresponsive, or simply slow, jobs is actually trying to do can be an invaluable clue. At times, it can point you in a useful direction long before enough performance data can be collected even to start an analysis.


    Featured Post

    How your wiki can always stay up-to-date

    Quip doubles as a “living” wiki and a project management tool that evolves with your organization. As you finish projects in Quip, the work remains, easily accessible to all team members, new and old.
    - Increase transparency
    - Onboard new hires faster
    - Access from mobile/offline

    Join & Write a Comment

    Suggested Solutions

    It Is not possible to enable LLDP in vSwitch(at least is not supported by VMware), so in this article we will enable this, and also go trough how to enabled CDP and how to get this information in vSwitches and also in vDS.
    When we have a dead host and we lose all connections to the ESXi, and we need to find a way to move all VMs from that dead ESXi host.
    Hi everyone! This is Experts Exchange customer support.  This quick video will show you how to change your primary email address.  If you have any questions, then please Write a Comment below!
    Excel styles will make formatting consistent and let you apply and change formatting faster. In this tutorial, you'll learn how to use Excel's built-in styles, how to modify styles, and how to create your own. You'll also learn how to use your custo…

    734 members asked questions and received personalized solutions in the past 7 days.

    Join the community of 500,000 technology professionals and ask your questions.

    Join & Ask a Question

    Need Help in Real-Time?

    Connect with top rated Experts

    21 Experts available now in Live!

    Get 1:1 Help Now