none
Exchange 2010 Performance Base-line RRS feed

  • Question

  • Good day

    I am looking to get a performance baseline of our organisations mailbox servers which currently has 4 databases with 200 users per DB on ecach Mailbox server, 4 Mailbox servers in total. The DB's are in a single DAG with each database having a passive copy on the other mailbox servers.

    Next year we are looking to add another 4000 mailboxes to the above configuration and want to then perform a another performance baseline that we can compare to the baseline taken prior to the extra 4000 mailboxes.

    How should I proceed to do this?

    1. Use ExBPA (performance report?)
    2. Run on ExBPA on each Mailbox server (how long per server?)
    3. How analyse the reports for performance problems such as high disk i/o etc

    Many thanks in advance.

    Patrick

    • Moved by Blake Mengotto Monday, June 18, 2012 5:04 PM exchange question (From:Operations Manager - Management Packs)
    Monday, June 18, 2012 9:29 AM

Answers

  • ExBPA is not really the tool for this, IMO.

    I would use Performance Monitor.  You might want to pick counters that you think are most useful from this guide:

    http://technet.microsoft.com/en-us/library/dd335215.aspx


    Ed Crowley MVP "There are seldom good technological solutions to behavioral problems."

    • Proposed as answer by Xiu Zhang Tuesday, June 19, 2012 8:10 AM
    • Marked as answer by Paddy831 Tuesday, June 19, 2012 8:27 AM
    Monday, June 18, 2012 9:47 PM
    Moderator
  • You don't want too many samples unless you have a huge amount of space and huge computational capability to analyze all of them.  I typically sample every 15 seconds over a workday.

    Ed Crowley MVP "There are seldom good technological solutions to behavioral problems."

    • Marked as answer by Paddy831 Monday, August 13, 2012 8:08 AM
    Saturday, August 11, 2012 6:24 AM
    Moderator

All replies

  • ExBPA is not really the tool for this, IMO.

    I would use Performance Monitor.  You might want to pick counters that you think are most useful from this guide:

    http://technet.microsoft.com/en-us/library/dd335215.aspx


    Ed Crowley MVP "There are seldom good technological solutions to behavioral problems."

    • Proposed as answer by Xiu Zhang Tuesday, June 19, 2012 8:10 AM
    • Marked as answer by Paddy831 Tuesday, June 19, 2012 8:27 AM
    Monday, June 18, 2012 9:47 PM
    Moderator
  • Thanks for the article Ed,

    When should I run the counters and how long? Should I run the counters for 2 hours during the busiest time?

    Much appreciated.

    Tuesday, June 19, 2012 8:29 AM
  • There is no one correct answer to your question.  It depends on what it is you're trying to determine.

    Ed Crowley MVP "There are seldom good technological solutions to behavioral problems."


    Tuesday, June 19, 2012 5:25 PM
    Moderator
  • Hi Ed

    Sorry I have not responded in so long I have been dodging this project for far too long. Just to refresh I would like to state what I would like to achieve.

    1. Performance Baseline of Exchange 2010 Mailbox Servers

    I plan to do this by installing all the relative Mailbox Counters in Perfmon.exe http://technet.microsoft.com/en-us/library/ff367871

    What I dont know understand is how long I should run the perfmon counters for? and at what periods? will the perfmon logging cause significant distress to the server performance?

    All your advice and help is appreciated.

    Thursday, August 9, 2012 9:42 AM
  • That document lists a lot of counters, and by monitoring all of them you'll likely have a very good baseline.  But they are likely more than you really need.  If you have the space and resources for all that monitoring, then it's fine to collect it all.


    Ed Crowley MVP "There are seldom good technological solutions to behavioral problems."

    Thursday, August 9, 2012 10:33 PM
    Moderator
  • For sizing you always capture peak load and peak concurrency. The more samples you get the better obviously.

    James Chong MCITP | EA | EMA; MCSE | M+, S+ Security+, Project+, ITIL msexchangetips.blogspot.com

    Friday, August 10, 2012 12:39 AM
  • You don't want too many samples unless you have a huge amount of space and huge computational capability to analyze all of them.  I typically sample every 15 seconds over a workday.

    Ed Crowley MVP "There are seldom good technological solutions to behavioral problems."

    • Marked as answer by Paddy831 Monday, August 13, 2012 8:08 AM
    Saturday, August 11, 2012 6:24 AM
    Moderator