We're investigating implementation options for SQL Server AlwaysOn.
AlwaysOn is dependant on Windows Clustering, and I have read several articles that state shared storage is no longer a requirement for clustering. I have also read several articles that state the opposite, or state that it only isn't a requirement for multisite clusters.
We do not plan on using multisite clustering. We would have two servers in the same room - one being a master server that takes all writes, and another backup server that we can use as a reporting server if available. On a failover, the secondary becomes the master until the next failover happens. We will also replicate some of our data off-site to two servers in a web farm. This is pretty standard from what I understand.
I'm on the software/database side of this project, so bear with me on these questions - I might need a lot of clarification:
- Is shared storage still a requirement for our proposed implementation of Windows Clustering?
- Why is shared storage a requirement, when I'm going to be replicating our SQL data to the other server anyway?
- Is shared storage only a requirement for system files? Can we have our MDFs/LDFs on directly attached storage? (we probably will use a SAN, but just curious)
Is it possible to make replication highly available, possibly by failing over the publishing responsibilities along with everything else? I have a solution in mind if we can't, but that would be really nice.