Clustering.

  • Hi Everyone,

    Firstly, I am not a DBA by training, but have become one as a result of need. (I am actually a web application developer that inherited the Server and SQL server instance - so while I can create just about anything in a query and have a "reasonable" handle of database design - when it comes to administering the database - I am very "green"

    The group I work for in the university currently has a 2000 (enterprise) installation and we are upgrading to 2005.

    I have been asked to investigate such things as clustering and replication.

    The management's understanding of these two seem to be a little different to mine and this ultimately leads to my questions:

    I thought clustering was to increase performance and "failover / duplication" was a "handy side-effect" of clustering. I also thought that replication as a stand-alone requirement was completely different to clustering and merely meant that there was more than one copy of the database(s) to enable a timely recovery from disaster (ie change the dns and you're back online)

    So, ultimately I am hoping that someone will be able to tell me if my "guesses" about the above are correct or not and perhaps also point me in the direction of a "plain english" article that might help, not only me but the management here to understand the technologies and what would be required to implement them.

    Thanks in adnvance for any help you might be able offer.


    Gavin Baumanis

    Smith and Wesson. The original point and click device.

  • Clustering, as far as SQL Server is concerned, is hardware fail over only.  The data is not duplicated and it doesn't help with performance. It's not load balancing. Basically, two physical servers are connected to the same disk space, probably on a SAN.  If one physical server goes down,the resources, including SQL Server and the disk, are failed over to the other physical box.

    Replication copies data from one server to another physical server.  I guess you could use this as a disaster recovery solution.  Probably log-shipping would work better than this for DR.  I am using such a solution for one of the large systems I support.

    Brian Knight has some articles on clustering you can find by searching this site.  Here is another one I found that looks promising as far as a good explanation: http://www.sql-server-performance.com/clustering_intro1.asp

     

    Hope this helps,

    Kathi

    Aunt Kathi Data Platform MVP
    Author of Expert T-SQL Window Functions
    Simple-Talk Editor

  • Hi Kathi,

    Thanks very much for the reply and advice.


    Gavin Baumanis

    Smith and Wesson. The original point and click device.

  • I forgot one thing since you will be using SQL Server 2005, look into Database Mirroring for DR.

    Aunt Kathi Data Platform MVP
    Author of Expert T-SQL Window Functions
    Simple-Talk Editor

  • Setting up clustering is not for the novice. You have to have MSFT approved servers (we found this out the hard way... our first ones were not and we had failovers quite often for no reason). The back end storage, multiple IP addresses, clustering, installing/configuing SQL Server...etc. Not to scare you off mind you, but, a guy off the street should not be setting this up.

  • Agreed; if the data is that important that 'the management' want to implement failover or replication, they should first invest in training or hiring a DBA, as either option has it's complications (speaking as someone who has had a fair amount of exprience with dealing with both).

    Assuming you are happy with inheriting the role of DBA, I would recommend you push for some training; it's not cheap but it's a hell of a lot cheaper than implementing a clustered or replicated solution that either does not work as expected or just isn't required...

    HTH

Viewing 6 posts - 1 through 5 (of 5 total)

You must be logged in to reply to this topic. Login to reply