I guess my view of the “statefulness” of MSCS is that you at least aren’t, from all appearances, tossing the file system and all applications to the wind, hoping that transaction logs and file system checks will keep you safe. I agree that an MSCS cluster is not totally and completely stateful, but it is certainly more stateful than VMware HA.

Not that there’s anything wrong with HA, per se; there’s just something about me that is uncomfortable with a guest OS rebooting, having to replay file system transactions, replaying application transactions, etc. I’ve seen, as I’m sure we all have, corrupted databases as a result.

For that reason, I’m looking forward to Continuous Availability, which I agree will bring a new level of statefulness that MSCS cannot even begin to touch.

I can’t claim to be a cluster expert either but my understanding is that MSCS is not a stateful solution. If the application needs to be restarted on the other node of the cluster the last string of application transactions that are not yet committed to disk get lost (as it is the case of a vm that needs to restart on another host). The difference is that with MSCS you only restart the app on the other node while with VMware HA you have to restart the whole vm (OS+app). My understanding is that there is no difference in terms of stateful/stateless capabilities.

In fact VMware Continuous Availability will bring a brand new level of HA into the game .. which is well above what MSCS can guarantee today.

Massimo.