System Design: Availability Patterns

1. Fail Over: Basically the fail over is kind of following:

But it is really more complex than above. If we double click a little, failover is:

The complete fail back is as follows:

2. Replication: There are two kind of replications:

• Active Replication: Based on PUSH model, we will discuss multiple approaches for this.
• Passive Replication: Based on PULL model. Whenever a node see that it does not have the requested data, it just read from peer and stores it locally. This strategy works well with timeout based caches.

Active Replications:

Master-Slave Replication:

Tree Replication:

 

Master - Master Replication:

Buddy Replication:

Now what happens if a node dies, say Node A crashes: