I am trying to think of different ways to build a highly available SQL instance, and have been mulling over different options. How much time difference is there during failover between a pretty standard 2-node active/passive SQL cluster and a database mirroring setup with high-safety and automatic failover?
The instance is a "nicer" method, as everything follows the failover (SQL agent jobs, etc), but in my experience an instance failover can take 2-3 minutes to actually take place. Additionally there are some shared single points of failure, with the primary one being the shared storage/SAN. If I go with a mirrored solution, I *think* the failover will be quicker, resulting in less DB downtime and there isn't the shared single point of failure (since the DB mirror doesn't have to have shared storage), but you lose the SQ agent jobs/etc being moved over after a failover. I might be missing some other benefits of a FCI that a mirrored option doesn't give you. The other big one I can think of is that the mirrored site doesn't share a DNS name, so some additional work has to occur after failover for clients to reach the mirrored DB, correct?
The main point of this thread is to understand what the time difference is between the two options. But I'd be happy to listen to any other factors that you think I have missed.