none
Unable to find the description of the resolved issue in KB

    Question

  • The following query runs for 10 mins on tbl with 1 mill records on sql server 2012 RTM, and for 1.5 min on sql server 2012 build 3128.

    dbcc freeproccache

    dbcc dropcleanbuffers

    go

    select *

    FROM "Demo Database (7-1)".dbo."CRONUS International Ltd_$Item" AS "Item" WITH(READUNCOMMITTED) 

    WHERE (ISNULL("Item"."No_",N'')>=N'')

    ORDER BY "No_" ASC

    go

    I am trying to pinpoint the build/hotfix that resolved the problem, and am unable to. Can you help?

    Wednesday, June 4, 2014 9:01 AM

Answers

All replies

  • I am trying to pinpoint the build/hotfix that resolved the problem

    Hello Jasminka,

    Which "hotfix" do you mean?

    Have you checked & compared the execution plan of the query on both SQL Server if you can see a difference?


    Olaf Helper

    [ Blog] [ Xing] [ MVP]


    Wednesday, June 4, 2014 9:29 AM
    Moderator
  • Hi Olaf,

    Well which hotfix is exactly the question. I know the issue exists on rtm but not on 3128 build, so I wanted to find out which build/hotfix resolved the issue, the description of resolved issues I've seen didn't match my issue (or not at first sight at least). Didn't compare execution plans, no. I assume that's what was fixed. I was just wondering if anyone can tell me what change/fix, and in which build (between RTM and 3128)  did the trick here.

    Thanks!

    Jasminka

    Wednesday, June 4, 2014 9:49 AM
  • I know the issue exists on rtm but not on 3128 build,

    Hello Jasminka,

    See http://sqlserverbuilds.blogspot.de/ : build 3128 is SP 1 + a hotfix: http://support.microsoft.com/kb/2793634/en-us


    Olaf Helper

    [ Blog] [ Xing] [ MVP]

    Wednesday, June 4, 2014 11:31 AM
    Moderator
  • The following query runs for 10 mins on tbl with 1 mill records on sql server 2012 RTM, and for 1.5 min on sql server 2012 build 3128.


    Significant differences in duration are usually due to different plans.  As Olaf suggested, comparing the plans is a good first step.  Assuming the plans are different, I wouldn't be too quick to assume the difference is related to the software build.  Different plans are often due to differences in the row count estimates used to generate the plan so make sure the same indexes and statistics exist in the databases and that stats are fully up-to-date.

    What is the purpose of the WHERE clause? It seems to me it will return all rows anyway.


    Dan Guzman, SQL Server MVP, http://www.dbdelta.com


    Wednesday, June 4, 2014 12:02 PM
    Moderator
  • Thanks for your replies.

    The tests are done on exactly same dataset, with statistics updated and cache cleared before each test. Tests are done multiple times (about 20 by now) so it is not a question of different row count, statistics, or anything else,

    results are consecutive. It is a build issue, i just needed to find out which build (from RTM to 3128) actually addressed this. Anyway, thanks for your effort.

    Tuesday, June 10, 2014 6:40 AM