locked
Mirroring Many Databases RRS feed

  • Question

  • We have an environment whereby we are mirroring over 20 databases on one server. We have not experienced any noticable performance issues or any other issues for that matter. However, I wonder if we may be stretching the limits of Database Mirroring or setting ourselves up for something really bad to happen.

     

    Our environment (read application) dictates that we create a new database for each client. So I do not see the number of databases in our environment decreasing or even stabillizing...only increasing.

     

    I suppose, my questions are:

    1. Are we heading towards issues having so many databases mirrored?
    2. Should we be looking at alternative ways of segragating our clients' data (schemas, other) rather than using separate databases?

    Any assistance and/or opinion is welcome.

     

    Tuesday, September 16, 2008 9:17 PM

All replies

  • It really depends on various factors like Memory, Bandwidth etc..Paul has blogged about this, please have a look at it.
    http://www.sqlskills.com/blogs/paul/2007/11/09/ConferenceQuestionsPotPourri4HowManyDatabasesCanYouReallyMirrorPerInstance.aspx

    - Deepak
    Thursday, September 18, 2008 5:27 AM
  • Thanks for the replies.

     

    Deepak...I have read that blog post in the past and you are correct...it really does "depend on various factors". Here are some additional details so you, and the rest of the community, have a clearer picture:

    • Memory: 4GB in both primary and mirror servers;
    • CPU: 2xDual Core 3GHz in both primary and mirror servers;
    • IO Bandwidth: This I do not know...sorry;
    • TL Generation: All DBs are used during normal business hours, some more heavily than others;
    • Network Bandwidth: There is a GBit connection between the primary and mirror servers.

    Our databases range in size from 1-40GB...nothing outrageous as far as I am concerned. My concern is the number of databases will continue to grow as we add more and more customers if we stay with our existing architecture. Do you think we should be looking at consolidating databases using schemas in order to save the stress on the mirroring infrastructure?

     

    Thanks.

    Thursday, September 18, 2008 1:55 PM
  • That is a valid point, however, if I had the Users connecting using a connection string that would limit which schema they had access to, then the two-part object name would not be required. Correct?

     

    [EDIT: Except for user-defined functions, of course.]

     

    Thursday, September 18, 2008 2:56 PM