Observe the following steps when considering node counts driven by high availability architectures:
- Always add one separate server for backups.
- Always allocate one server for a logical or physical replica.
- For automated failover, allocate the following:
- An additional small VM / node to act as a voter
- OR a fully qualified replica
- For every active data center beyond the first two, allocate one replica.
- If non-local data access latency is a concern, allocate the following:
- An additional replica in the primary location
- An additional replica in each location for symmetric clusters