Cross Data Center Replica

Hi there,

we would like to span a MongoDB cluster across multiple data centers for Geo-HA.

The three data centers we have in use have the following latencies:

A <> B: 5 ms
A <> C: 25 ms
B <> C: 25 ms

The database is used with real-time applications (user frontend…) and therefore we cannot have any long latencies with reads and writes.

  1. What impact do network latencies (5 ms, 25 ms) have on read and write operations within a cluster?
  2. What would the effect be if C is only an Arbiter (no replication)?