Planning Analytics

 View Only
  • 1.  PAW - REDUCE THE SIZE OF THE FEEDER FILE

    Posted Tue June 06, 2023 02:00 PM
    Hello,
    I have a TM1 cube that has a very large Feeder file (+-25GB). I would like to reduce the size of this file.

    Are there any tips for figuring out which rule/feeder is using this and how to narrow it down?

    How to find the correct rule I need to feed?

    Any help will be appreciated.

    Thanks!


  • 2.  RE: PAW - REDUCE THE SIZE OF THE FEEDER FILE

    Posted Wed June 07, 2023 02:08 AM

    Hi,

    1. Have you already reviewed your rules for overfeeding?
    2. Do you use persistent feeders?
    3. Can you read, these articles first?:
      1. TM1 Server Feeders Guidelines and Best Practices (ibm.com)
      2. Over Feeding in TM1 - What is it and How to Avoid it - Exploring TM1
      3. How to find over feeding in your TM1 model - Cubewise CODE

    Regards,

    R. Klingiert



    ------------------------------
    Robert Klingiert
    ------------------------------



  • 3.  RE: PAW - REDUCE THE SIZE OF THE FEEDER FILE

    Posted Fri June 09, 2023 08:23 AM

    Hello @Robert Klingiert

    Thanks for the links! It will be very helpful.

    But I'm looking for a way to identify which Rule/Feeder is taking up a lot of memory space.

    I have a lot of rules/feeders in my cube, I would like to identify which specific rule/feeder I can improve.

    Do you think using Performance Modeler->Generate report is a good approach?

    Thanks!



    ------------------------------
    Paulo Klein
    ------------------------------



  • 4.  RE: PAW - REDUCE THE SIZE OF THE FEEDER FILE

    Posted Fri June 09, 2023 08:59 AM

    Hi Paulo,

    Let's use the way proposed by Andre, but wie shall analyse data in }StatsByRule CUBE:

    This is my demo data:

    a) First, you have to turn on the statisitc: }StatsByRule - IBM Documentation.: "To enable the storage of statistics for a rule, you must set the Rule_Stats property for the associated cube to YES in the }CubeProperties control cube."
    b) Next you have to recalculate your rules. You can do it restarting your TM1 Server (if you have persitent feeders, then backup and delete all *.feeders files from your model directory)

    c) Run any client (Architect, PerformanceModeler, PAW) and find }StatsByRule cube in Control Objects (You have to check an option i the menu of Architect or PM):
    d) create any suitable view for you analysis, for example:

    e) As you can see, I sorted the data by the "Total run" column. You can easily indicate which rule is a feeder and compare the feeder with the corresponding rule:

    Try this.

    Regards.



    ------------------------------
    Robert Klingiert
    ------------------------------



  • 5.  RE: PAW - REDUCE THE SIZE OF THE FEEDER FILE

    Posted Mon June 12, 2023 01:43 AM

    Hi,

    Ad:"b) Next you have to recalculate your rules. You can do it restarting your TM1 Server (if you have persitent feeders, then backup and delete all *.feeders files from your model directory)"

    A little explanation, to be clear:

    a) first stop the TM1 server,
    b) backup and delete all *.feeders files from your model directory,
    c) start the TM1 server.

    Regards,



    ------------------------------
    Robert Klingiert
    ------------------------------



  • 6.  RE: PAW - REDUCE THE SIZE OF THE FEEDER FILE

    Posted Thu June 08, 2023 03:50 AM

    Hi Paulo,

    Your question touches on the holy grail of Planning Analytics/TM1; it is a complex question that can't really be answered here. However, I will attempt to give a short answer that will hopefully, point you in the right direction. But it also assumes that you have a good understanding of feeders and what they are used for [ here some info just in case: TM1 Server Feeders Guidelines and Best Practices (ibm.com) ].

    • In most cases, setting feeders correctly requires time and focus and that "investment" is best done when your write the rules.
    • Trying to identify overfeeding after the fact is more time consuming.
    • There is no "sliver bullet" to fix overfeeding. Even just determining overfeeding is not that straight forward as a certain amount of overfeeding is often unavoidable.

    My recommendations are the following:

    • Start by having a look at the }StatsByCube
      • The "Number of fed cells" for a cube will be in direct relation with the size of the feeder file (i.e. created when PersistenFeeders=T in the TM1.cfg)
      • The ratio between "Number of fed cells" to "Number of populated Numeric Cells" can give an indication if there is massive overfeeding or not. But this depends on your model; you must have a good understanding of you model to make that call.
    • Method/Technique to check for overfeeding
      • In the book "IBM Cognos TM1: The Official Guide", there is the description of how to implement one method on how to check for overfeeding (page 146  through 149). Currently this book in paperback is less than $5 (check on the internet) and has some good information; might be a good investment 😊
      • Alternative method: Troubleshooting Feeders in TM1 with Emojis 💪 - Cubewise CODE
    • Once armed with information from the above, review the relevant rules and especially the relevant feeders and make changes where appropriate. Again, there is no sliver bullet, this will be tedious work, but with one of the method described above, you should be able to identify the "low hanging fruits" (i.e. the feeder statements with highest potential) and reduce the overfeeding.
    • Benefits:
      • Your feeder file should shrink
      • Start up time, will be shorter.
      • Some performance improvement should be expected.

    Keep in mind that IBM and its business partners offer health checks and performance tuning services in case you do not have the knowledge/skills and/or the bandwidth to tackle this on your own.

    Best Regards,



    ------------------------------
    Andre
    ------------------------------