IBM i Global

 View Only
  • 1.  IBM Navigator for i - Performance Data Investigator - usage questions

    Posted Thu December 05, 2024 04:20 AM

    First is a general question, where to find help on the PDI section of Navigator for i?  

    Here are my questions, can someone point me where I should be going to find the answers please.....

    1. There is a PDI report called Page Faults for One Pool. The graph that comes up appears to just show the total page faults for the system. If I click to show the data in table form, I can filter on the pool I want a report on. I then click on the chart, it doesn't graph the data I filtered in the table, it just shows a graph of the total of all pools. How do I get that graph please? 
    2. One of the metrics in that graph is 'Pending Faults per second'. How can I get an explanation for what that means?
      • Can someone provide a general description of 'Pending Faults per second', whether they are good or bad and how to assess their impact on the pool performance.

    Many thanks



    ------------------------------
    Edward Yeates
    ------------------------------


  • 2.  RE: IBM Navigator for i - Performance Data Investigator - usage questions

    Posted Thu December 05, 2024 10:41 AM

    Edward

    I can't get that graph to work either but..this graph does the trick



    After you select the above graph, scroll down until you get to this section for Page Faults


    Once you select which pools you want ( I chose deselect all then selected shrpool1 for instance) you will get this graph



    As far as the pending faults are concerned - this comes from the wait bucket document itself

    You can read more about wait buckets here -> Job Watcher Wait Buckets and "enums"

    I've always 'dismissed' the pending faults  and just focused on actual faults - just one man's perspective anyway

    Also - full disclosure - I don't generally use PDI for performance analysis, so others may have a better understanding/work around for the graph you mentioned. 

    Hope this helps and good luck - Rich



    ------------------------------
    Rich Malloy
    ------------------------------



  • 3.  RE: IBM Navigator for i - Performance Data Investigator - usage questions

    Posted Fri December 06, 2024 01:50 AM
    Edited by Satid S Fri December 06, 2024 01:57 AM

    Dear edward

    >>>> where to find help on the PDI section of Navigator for i <<<<

    An ex-IBMer expert named Dawn May posted many blog (as well as presentation) about using PDI in  https://techchannel.com/i-can-blog  and https://dawnmayi.com/ican-blog-archive/ . (I have accumulated several now). A sample here: iCan Blog Archive: View Memory Pools and Faulting With the Performance Data Investigator at https://dawnmayi.com/2013/11/11/view-memory-pools-and-faulting-with-the-performance-data-investigator/      

    So, I suggest you include her name in your Google search to increase the chance of finding helpful information about using PDI tool.   I also would like to add that many PDI charts are quite easy to interpret (especially for one usually involved in IBM i performance concept).    Also ask your questions here and I can help answer them.

    >>>> One of the metrics in that graph is 'Pending Faults per second'. How can I get an explanation for what that means?  <<<<

    Pending fault happens when multiple jobs want to access data residing on the same disk page(s) at the same time.  The page faulting caused by the first job is counted as normal fault, the remaining jobs are counted as pending fault.     In my experience, running very many data query jobs in a memory pool is a usual (not always) cause of high count of pending fault rate in that memory pool,  In term of performance interpretation, you focus only on the normal faulting rate.   

    To use PDI chart on memory faulting rate overview sensibly, you should also look at the chart on Wait Overview of the same date-time period.   You want to see if the high memory faulting period coincides with high Disk Page Fault Time in Wait Overview chart of the same period or not.  If not, you do not worry about high memory faulting yet.    Another aspect is to compare Memory Faulting Rate Overview chart with Disk Throughput Overview for Disk Pools chart of the same date-time period.  If you see disk response time (disk service time + disk wait time) that is consistently lower than 5 msec. during high memory faulting period, then you do not worry about high memory faulting yet.

    One thing to keep in mind.  If your system has a lot of concurrent active jobs running Query/SQL without much useful indexes created on the tables being accessed, this is a frequent cause of high memory faulting rate.  Using Db2i index advisor to create as many useful indexes as possible can substantially to significantly reduce high memory faulting rate. 

    Hope this cast some light for you.  Feel free to ask more if you wish.



    ------------------------------
    Satid S
    ------------------------------



  • 4.  RE: IBM Navigator for i - Performance Data Investigator - usage questions

    Posted Fri December 06, 2024 09:07 AM

    Start with the Page Faults by Pool graph.  If you have the HTTP groups updated in 2024, you will be required to select ONE pool before drilling down to the "Page Faults for One Pool" graph.  This will show you only the data for that one pool, over time.

    page faults by pool
    showing previously selected pool over time for this one pool


    ------------------------------
    Lora R Powell

    IBM i – Systems
    IBM Navigator for i & Performance on the web (PDI)
    IBM i 7.5 Memo to Users
    IBM Power Systems Development

    Office: 30-2/ C204

    Do the next right thing
    Let's GO
    ------------------------------