We are planning a new SQL cluster. We intend to have the two servers separated by 100 meters. Will the cluster’s shared disk array be a single point of failure for us?

A cluster is designed to provide fault tolerance for the physical servers, the operating system, and the application software, but not for data. Fault tolerance for data should be handled by a fault tolerant shared disk array, log shipping, or database mirroring (in SQL Server 2005).

So in a sense, the shared disk array is a single point of failure should the entire array fail. But if the shared array is well designed and employs fault tolerance features, it should be very reliable. This is a good reason why you should not skimp on spending money on a shared disk array for a SQL Server database cluster.

If you intend to separate the two nodes of the cluster by 100 meters, you will have no choice but to use Fibre Channel connections to the shared array, as SCSI is not supported for such a long distance.


No comments yet... Be the first to leave a reply!