Distributed Systems

Chain replication

Replicas in a linear chain; writes flow head→tail, reads from tail.


In plain terms

van Renesse & Schneider, 2004. Strong consistency with simpler failure handling than Paxos. Used in Microsoft's Azure storage and FoundationDB.

Origin

Robbert van Renesse and Fred Schneider, "Chain Replication for Supporting High Throughput and Availability," OSDI 2004. Strong consistency with simpler failure handling than Paxos.

Where it shows up in production
  • Microsoft Azure Storage Uses a chain-replication variant for the partition-layer log.
  • FoundationDB Tiered architecture with chain-replicated transaction logs.
  • CRAQ extension Reads-from-any-node variant; allows scaling reads across the chain.
Sources & further reading
Found this useful?