What are all the pros and cons of multi-leader vs leaderless replication in databases?
Here are some I can think of:
- In multi-leader, client only writes to one leader so there may be lower latency
- In leaderless, you can generally tolerate a greater number of failures because as long as there are w nodes available for writes and r nodes for reads then the system can continue functioning
What are some of the other benefits / drawbacks of each approach?