I was asked this question today. Why did Microsoft pick “16” as the maximum number of servers that you can have in a DAG? Having 16 servers in a DAG means that you can have a maximum of 16 copies of a database (one active and 15 passive). Isn’t it a bit too much to have 15 passive copies of a database? How can you justify this figure?
These were the questions that bugged an exchange admin that I met. To make it clear, the Exchange team didn’t go for a “magic” number for the maximum number of servers in a DAG. The limit is imposed on the DAG by the underlying Windows failover clustering, on which DAG is based.
The Windows failover cluster can only support a maximum of 16 nodes. This limitation kind of flows into DAG. So, though a DAG can only have a maximum of 16 servers, the limit actually comes from the Windows failover cluster.
There you go, the story behind the magic number