We are looking at the possiblity of using NT Clusters to improve our reliablity. The problem is that while clustering is great in theory ,if not implemented exactly right it can be LESS reliable than a single machine.

Does anyone here know of anyone who has used NT clustering and gotten greatly INCREASED reliablity.


One thing we are thinking of is a high latency factor so that the cluster machine is at least 15 minutes behind the one it is backing up. That way problems which crash a machine won't automatically crash every mahcine in the cluster. But it might cause other problems.

I guess there are two problems here:
Can clustering increase reliablity?

What kind of clustering increases reliablity, if any does?

Obviously if go this route we will hire a consultant to handle this.


But question remains -- is clustering the best route to go.

Another possibilty is to keep a backup machine with nothing but the OS on it. In case of disaster we then use normal restore procedures to retore to the backup machine. If clustering is reliable this is slower recovery. But it will be faster than unreliablle clustering -- since if an entire cluster goes down it will take longer to get back up that a single machine.

Any thoughts in this area will be greately appeciated