A very simple (yet effective) replication system
This is the experimental branch for λDB (BugSense's internal database). Expect to see lots of stuff changing all the time, formats being broken etc.
Motto: Less is more
How it works
Read more at BugSense's post:
Every node is the same. When the Master node goes down, the last-to-data-written replication node, is promoted to Master. When the Master node comes back, it is just a new replication node.
- Simple/fast implementation (a.k.a less bugs)
- Minimal downtime of the service
- Best ration consistency/speed
- NO guarandeed 100% consistency (there are more complicated algorithms and way better designed. Read: PAXOS).