Skip to content

SYSTEM Cited by 2 sources

Amazon Dynamo

Amazon Dynamo is the internal Amazon key-value storage system described in the DeCandia et al. SIGOPS 2007 paper. It is not the same system as Amazon DynamoDB (the 2012 public AWS service that shares the name and lineage). Dynamo-the-paper established the template that Cassandra, Riak, Voldemort, and many others followed:

  • Consistent hashing for key partitioning across nodes.
  • Configurable replication — N replicas per key with tunable R, W quorums.
  • Versioned objects + vector clocks for concurrent-update reconciliation.
  • Gossip-based membership + failure detection — peer-to-peer, no central coordinator.
  • Merkle-tree anti-entropy for replica reconciliation.
  • Hinted handoff for temporary-failure writes.

Why it uses gossip

From sources/2023-07-16-highscalability-gossip-protocol-explained:

"Amazon Dynamo employs the gossip protocol for failure detection, and keeping track of node membership."

Gossip was the "no central point of coordination" mechanism that let Dynamo's 2007 design claim operational properties ("always writable", availability-biased) that the centralized alternatives (single master, Chubby/ZK-coordinated) could not match at Amazon's scale.

Wiki scope

This page is a stub anchored to named use in secondary-source explainers. For the authoritative Dynamo material, cite the SIGOPS paper directly and Werner Vogels's write-ups on allthingsdistributed.com. Downstream open-source descendants — Cassandra, Riak, Voldemort — have their own pages for production-operator documentation.

Seen in

Last updated · 319 distilled / 1,201 read