SQL Server Performance Forum – Threads Archive

Cluster & NLB

I#%92ve been researching SQL Server clustering for my company, and your articles on sql-server-performance.com are very informative. However, I was hoping you could clear up some unknowns for me. If I#%92m to do active/active 2 node clustering, do I have to, and how would I set up special memory restriction for the original sql server instance. I read that if a node fails 2 instances of SQL Server will need to run on the same machine. The reason why I want to run active/active is because I want to utilize both machines. My company can#%92t afford having a machine being idle. What are your thoughts on load balancing, and is it possible for SQL Server. I read this link :http://www.microsoft.com/technet/prodtechnol/sql/2000/deploy/hasog04.mspx concerning Network Load Balancing. Is this an effective solution? Any feed backs and/or additional resources you may have is greatly appreciated.

SQL Server requires operating system resources to manage its databases. The most important of these resources is memory. To ensure adequate response times to clients during a failover and to ensure SQL Server Failover Support functions as expected, a review of memory requirements and anticipated server usage must be performed. Dynamic allocation of memory is important during a failover because it allows SQL Server to adjust to its new environment by optimizing its memory usage. In an active/active configuration, both servers running SQL Server will optimize their memory usage automatically during a failover. If the SQL Server virtual server on node 1 fails over to node 2, the physical memory requirements for all services on node 2 will now be 160 percent. Both SQL Servers virtual servers will reallocate their memory requirements dynamically to the available memory, which is 50 percent, the amount of memory originally available to SQL Server. By viewing Windows NT Performance Monitor, you may decide to stop one or both of the additional services gaining 15 percent more memory for each service that is stopped. Satya SKJ
Moderator
http://www.SQL-Server-Performance.Com/forum
This posting is provided â€œAS ISâ€ with no rights for the sake of knowledge sharing.

If you want you can set a memory limit for each instance but you could let SQL Server handle it automatically too. About loadbalancing it’s often not something you can apply afterwards. It’s a something you need to think about during the design of each database and application. There are many different ways to achive it. For example you could have one master database that is read/write on a failover cluster and a number of single servers that keep a read only copy of the database. The application will write to the master but read from the others. It could be a simple round robin selection when the application needs to read data. Depending on the application and database the master could replicate the data with snapshot or transactional replication or you could detach/copy/attach the databases during night.

Hi! Here are my 2 cents for MSCS and NLB If you are planning to use MSCS for high availability and NLB for load balancing together then I don’t think it is supported by Microsoft Thanks,

]]>