Results 1 to 6 of 6
  1. #1
    Join Date
    Apr 2012
    Posts
    156

    Unanswered: BLU - Proof of Concepts

    I pulled this out of another discussion that was starting as it is more of its own topic. Where I am hoping BLU delivers is in the BI environment where when building aggregates/cubes etc. I am looking at it for SAP BI environments as an alternative to HANA, if we can get 95% of the performance of HANA with 1/3 the cost it will be a no brainer. I am curious has anyone done any performance testing on BLU? If so how was the migration? How long did the migration take (yes I know depends, but say migrating 10 TB's). I know it is really new so this would probably just be proof of concepts. I am trying to estimate resources (staff hours) needed for a proof of concept and would like to hear some real world experience (other than marketing :-) )

  2. #2
    Join Date
    Mar 2003
    Posts
    280
    Quote Originally Posted by azready View Post
    I pulled this out of another discussion that was starting as it is more of its own topic. Where I am hoping BLU delivers is in the BI environment where when building aggregates/cubes etc. I am looking at it for SAP BI environments as an alternative to HANA, if we can get 95% of the performance of HANA with 1/3 the cost it will be a no brainer. I am curious has anyone done any performance testing on BLU? If so how was the migration? How long did the migration take (yes I know depends, but say migrating 10 TB's). I know it is really new so this would probably just be proof of concepts. I am trying to estimate resources (staff hours) needed for a proof of concept and would like to hear some real world experience (other than marketing :-) )
    Probably considered marketing, but you may find some useful info in there:

    http://public.dhe.ibm.com/common/ssi...L14378USEN.PDF
    --
    Lennart

  3. #3
    Join Date
    Mar 2012
    Posts
    16
    Well we are preparing but .... it is currently not possible for DPF databases so now we must think about migriting to single node .....
    For existing environment you'll need to upgrade to 10.5FP1 and patch/upgrade you SAP applications according to note:
    http://service.sap.com/sap/support/notes/1889656
    Then it will be easy , apply recommended settings for BLU and convert your tables to columnar ones (tablespaces) using DB6CONV report (ADMIN_MOVE_TABLE procedure behind) with compression.If your 10TB db is uncompressed you can expect shrinking it to 2-3 TB in the end.
    Theoretically applying BLU shouldn't touch any BI logic and configuration - all work will be made by Basis and DBA staff - but knowing SAP it won't be as easy as expected

  4. #4
    Join Date
    Sep 2007
    Location
    Moscow
    Posts
    41
    Have little internal POC on cloned SAP BW 9.7 database upgraded to 10.5

    The main goal was to check BLU 's "in-memory optimized" feature for star query's over BW InfoCube tables (/BIC/FZ*) and some SELECT's over BW active ODS tables (/BIC/AZ*)
    Existing fact and dimension tables was copyed as "columns"-tables and some query's from production database snapshot run to compare with existing row-tables (adaptive compressed MDC).

    My results is an ambiguous.
    Some query run 2-3 times faster on column-tables, some is 2-3 times slowly than on row-tables.
    In any case BLU is very hungry for CPU and RAM.

    After our test's we decide what BLU is not needed for our BW right now

  5. #5
    Join Date
    May 2003
    Location
    USA
    Posts
    5,737
    Quote Originally Posted by mitrofun View Post
    My results is an ambiguous.
    Some query run 2-3 times faster on column-tables, some is 2-3 times slowly than on row-tables.
    In any case BLU is very hungry for CPU and RAM.

    After our test's we decide what BLU is not needed for our BW right now
    BLU is designed to use as much CPU and RAM as possible with intra-partition parallelism. That is one reason why it probably will not work well in DPF (even if made compatible) since DPF partitions typically only have 1 or 2 CPU cores per partition (if you have more CPU cores than that, additional partitions should be created for more parallelism). IBM normally recommends that intra-partition parallelism be shut off for DPF (it needs to be on for column based tables).
    M. A. Feldman
    IBM Certified DBA on DB2 for Linux, UNIX, and Windows
    IBM Certified DBA on DB2 for z/OS and OS/390

  6. #6
    Join Date
    Apr 2012
    Posts
    156
    Thanks for the feed back, what I am looking at is using BLU as an alternative to DPF. I'll post how our testing goes.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •