Hi there. I've read the Microsoft tested limits for Availability Groups (10 groups / 100 databases), but it's not clear how those apply (e.g. 1 group of 100 vs 100 groups of 1).
We have about 87 production databases, averaging 5 GB in size. They are not huge (largest is 115 GB at the moment) and none of them are high transaction databases (think TFS, SharePoint, etc). With only 10 groups, depending on how we carve up the groups, we may need to move databases between groups later, and we would prefer to not have to change connection strings, etc.
One idea we had was to have a group per application. These 87 databases support about 40 unique apps, so we could have 40 groups, with an average of ~2 databases per group. This would allow us to change group properties (aync vs sync, number of nodes, etc) without having to move databases between groups that have the designated properties.
Is the overhead more per group or per database? If it's per database, then I assume we could have 40 groups without a problem. If there is a lot of threading / monitoring overhead per group, then I assume 10 groups is the max.
This, of course, assumes our hardware / network capacity is sufficient for the overall load.
Thanks for any info.
Sam