Distributed Systems

Hinted handoff

When a replica is down, a temporary node holds its writes and replays them later.


In plain terms

Dynamo-style availability technique. Handoffs are time-bounded; if the replica stays down too long, anti-entropy takes over.

Origin

Dynamo paper (DeCandia et al., SOSP 2007) named the technique. The idea predates Dynamo — write-behind buffers in older systems served the same role.

Where it shows up in production
  • Cassandra A coordinator stores hints for a down replica for up to 3 hours by default; replays when the node returns.
  • Riak Same pattern — hints buffered locally and replayed asynchronously.
Sources & further reading
Found this useful?