none
any distance limited for failover clustering solution?

    Question

  • Could I implement a failover cluster solution on the two DBs which are based in two different cities?
    Possible?

    Wednesday, January 02, 2008 7:01 AM

Answers

  • You can do a geographically dispersed cluster, but it would be very expensive (two SANs, fiber, etc.) and the whole solution - as if it was a local one - would need to be in the Windows Server Catlog as a geographically dispersed cluster. See http://support.microsoft.com/kb/28074. There is a distance limitation which is basically fiber which has about a 100 mile limitation - it has nothing to do with SQL but everything to do with physics.

     

    If you need disaster recovery, log shipping or database mirroring are better choices. Replication is not a great D/R strategy if you need the whole DB.

    Wednesday, January 02, 2008 4:24 PM
  • For any DR solution to deliver a zero data loss guarantee, as clustering does and synchronous mirroring does, you cannot get away from some form of round trip to make sure that the principal doesn't get ahead of the copy.

    For DB mirroring, that takes the form of logic internal to SQL.

    For geographically dispersed clusters, the storage hardware doing the replication is doing the same thing under the covers.  That's why vendors like EMC have limits to how far they recommend separating synchronous replication pairs.

     

    Thursday, January 03, 2008 7:52 PM
    Moderator

All replies

  • As far as i know , the failover cluster solution is based on the following core components :-

     

    - Shared disk array . How will two computers in two differnet cities share it ??

    - Heartbeat which is a high speed reliable private network among the differnet nodes of the cluster to check the validity and status of nodes.  How far reliable will be a connection between two sites in two differnt cities ??

     

    I think database mirroring can be applied over WAN , but how reliable and fast will be your link ???  Microsoft even makes no recommendations about how reliable your WAN connnection should be .

     

     

    Wednesday, January 02, 2008 3:41 PM
  • Thanks for your reply.

    But if DB mirroring , every application request needs handshaking with 2 database(waiting for the whole acknowledgement with 2 db ), more or less, the performance  will  decrease.
    This is why i  abandon the Mirroring  solution.


    Now i am comparing Log shipping and Replication.

    How do u think about these 2?
    which will be better?
     
    Wednesday, January 02, 2008 3:50 PM
  • You can do a geographically dispersed cluster, but it would be very expensive (two SANs, fiber, etc.) and the whole solution - as if it was a local one - would need to be in the Windows Server Catlog as a geographically dispersed cluster. See http://support.microsoft.com/kb/28074. There is a distance limitation which is basically fiber which has about a 100 mile limitation - it has nothing to do with SQL but everything to do with physics.

     

    If you need disaster recovery, log shipping or database mirroring are better choices. Replication is not a great D/R strategy if you need the whole DB.

    Wednesday, January 02, 2008 4:24 PM
  • For any DR solution to deliver a zero data loss guarantee, as clustering does and synchronous mirroring does, you cannot get away from some form of round trip to make sure that the principal doesn't get ahead of the copy.

    For DB mirroring, that takes the form of logic internal to SQL.

    For geographically dispersed clusters, the storage hardware doing the replication is doing the same thing under the covers.  That's why vendors like EMC have limits to how far they recommend separating synchronous replication pairs.

     

    Thursday, January 03, 2008 7:52 PM
    Moderator