First thing, I would make sure whether the cluster is part of the requirement or not. There are ways to achieve continuous availability, disaster resilience, fault tolerance, and no-single-point-of-failure without using a cluster.
In my opinion, you need to be looking at vendor responses for indications on whether
* the backup site is "hot" (always live) and whether it's an automated or manual switch-over
* what kind of track record they have in the industry -- how many years have they been doing this and will they give you customer references you can call to ask for references
* what kind of part replacement turnaround is there? Top of the line continuous availability systems are monitored continuously by the vendor and broken partsare automatically noted and sent next-day delivery
* on a demo system that is running an application (hopefully somewhat similar to yours) can you reach in and pull a board (or disconnect a cable, etc.) without causing harm to the system as a whole. If you can't, then whatever side effects you notice are what will happen if that part dies -- there's your single point of failure.
* You'll want to know the average availability time for the server. Ask how they calculate it. There's a big difference between 99%, 99.9%, 99.99%, etc.
* In addition, you'll want standard server stuff -- CPU speed, memory, backup options (and speed of the backup). Give them an idea of what your application is like and tell them what the maximum load is likely to be.
* You'll want to know what kind of network/communication between the two servers (main server and backup).
* What sort of monitoring tools are available? Are they included in the price?
* Think about what kind of operation system you're willing to support. Does it need to be Windows? Unix?
