Riak Makes Applications "Completely Available"

To be truly “highly available”, your database must be available not just for reads but also for writes. If your database relies on a single master node to handle all of your application’s write requests, a crash will result in data loss until your failover instance comes online. The inability to write to your database means lost revenue and lost users.

Riak redefines commonly accepted levels of availability. Applications built on Riak are completely available; they are perpetually online for both reads and writes.

Riak’s shared-nothing architecture makes this possible. The shared-nothing architecture enables nodes in a Riak cluster to coordinate requests via asynchronous messaging. So, if a node in a Riak cluster crashes or hangs, the other nodes can handle the failed node’s requests, operating without clients noticing any disruption.

If a node becomes unreachable for any reason, Riak compensates using hinted-handoff. This means that neighbors of the unreachable node will receive the messages that the failed node would otherwise have handled, at the same time ensuring that features such as replication are not degraded in the window before the failed node is repaired or replaced.

Contact us or read more about how Riak ensures high availability.