Results 1 to 4 of 4
  1. #1
    Join Date
    Oct 2010
    Posts
    94

    Unanswered: BCU reusing SAN switch

    Hello,

    Quick questions -

    1) Our BCU environment is reported to have one SAN switch connecting the fiber from independent nodes in the cluster to their respective LUNs. Does this cause an IO bottleneck at all?

    2) Also, we have a fairly log-intensive ETL process (massive logged inserts, updates and deletes) but the same LUNs host both our data and log files for a given node in the cluster. Also the LUNs are all RAID-5. Does RAID-10 significantly speed things up for log writes?

    THANKS AS ALWAYS!!!

    - G

  2. #2
    Join Date
    Apr 2012
    Posts
    1,035
    Provided Answers: 18
    I cannot comment about your SAN switch, but your SAN folks will have diagnostics to answer that.

    Always wise to have separate LUNs for active transaction-logs - but more specifically on different physical resources (different arrays, different spindles). Unfortunately costs or fixed procedures usually prevent DBAs from achieving this. Plus many SAN folks just say 'no longer necessary'.

    If your SAN offers different tiering (e.g some LUNs at Raid 10 , others at higher Raid levels) for an acceptable annual cost to the business, the best thing to do is *measure* the difference of putting the active txn-logs on more performant LUNs. That's what matters.

    However, If your physical model could somehow significantly reduce or eliminate the logged-inserts during ETL (in favour of load) you would gain *far* more...

  3. #3
    Join Date
    Oct 2010
    Posts
    94
    What about switching the tablespace to NOT LOGGED? If we were to do that for daily incremental loads to the ODS and one of the loads were to fail, would we have to just rerun the load to get the data back, or would we have to literally drop and recreate the table before rerunning the batch?

  4. #4
    Join Date
    May 2003
    Location
    USA
    Posts
    5,737
    Quote Originally Posted by getback0 View Post
    What about switching the tablespace to NOT LOGGED? If we were to do that for daily incremental loads to the ODS and one of the loads were to fail, would we have to just rerun the load to get the data back, or would we have to literally drop and recreate the table before rerunning the batch?
    If you use the LOAD command with NONRECOVERABLE option, then there will be no logging (except for some minor logging that the LOAD has started, etc) and it will not put the tablespace in backup pending mode. See the Command Reference manual for more info on NONRECOVERABLE.

    With regard to BCU, the objective is create an environment were there is no one single obvious bottleneck between CPU, I/O, Network, etc (they are approximately equally constrained). Whether you have a specific bottleneck that much worse than the others, depends on your workload and on your complete hardware configuration.
    M. A. Feldman
    IBM Certified DBA on DB2 for Linux, UNIX, and Windows
    IBM Certified DBA on DB2 for z/OS and OS/390

Posting Permissions

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