SQL Server Performance

Basic doubts

Discussion in 'SQL Server 2005 Clustering' started by filson, Dec 24, 2006.

  1. filson New Member

    I am a new guy to this subject.SO I have basic conceptual doubts

    1)If we are using load balancing cluster.Database physical file will be same for all nodes or not?
    2)if Database physical file is different for nodes how the node sychronising data changes to other nodes
    3)Whether that synchronization is transaction by trasaction in seconds like replication
    4)suppose I suddenly pull out one node from cluster what will happen to the transaction.
    suppose it is a banking transaction debit transaction over when i am pulling out node.
    credit transaction will be done in any other node automatically.because in application i am mentioning data source name as cluster only na

    Thanks in Advance
    Filson
  2. Argyle New Member

    There exist no such thing as a load balancing cluster when it comes to SQL Server. The cluster option in SQL Server is about high availiblity and not load balancing. There is no data synchronization going on.

    A virtual SQL server instance in a failover cluster environment is seperate from any other instanses on that cluster with its own network name and ip and physical disks and databases.

    During a failover on such a cluster from one node to the other, the SQL services will be stopped on one node and started on the other. They are not running at the same time on multiple servers. Applications would have to deal with a failover the same way they would have to deal with a stop/start of the SQL services on a stand alone server. It's the same process that occur, just that the services start on another machine.

    If you want to achive some form of load balancing, you have to program your application for it. For example you could have your database in read/write mode on a failover cluster (for high availiblity) and then you could logship or restore databases in read only mode on multiple servers. Your application could then use some form of round robin algorithm to choose which server to read from. Any write activity would go to the central database on the cluster. You could even load balance the read only databases with Microsoft NLB.
  3. filson New Member


    Thanks Argyle,
    But still I have some doubt.In so many articles I read about Load balancing cluster.One of them is given below
    http://searchsqlserver.techtarget.com/tip/0,289483,sid87_gci1057987,00.html?bucket=ETA&topic=301327

    If you have some time for me go through it and clear for me.Why I am confusing that if load balancing is not there means what is significance of Active/Active cluster.Fail over will be supported by Active/Passive na

    Thanks & A Happy New Year
    Filson
  4. satya Moderator

  5. Argyle New Member

    Active/Active is just an option where you utilize more nodes in a cluster instead of having a node sitting idle in passive mode most of the time. To go active/active you would have some SQL instances running on one node and the rest on the other. Note that running active/active increases your licensing costs. Each active node needs licenses, a passive do not.

Share This Page