Data Replication with Apache Accord
summary
Apache Accord is presented as the new replication algorithm used in Apache Cassandra that enables ACID transactions by using leaderless replication where replicas coordinate via quorums. The talk explains how Accord builds on distributed-consensus work (Paxos, Raft, and Egalitarian Paxos), covers implementation and validation challenges (implementations in Go/Java, model checking, Jepsen testing), and demonstrates Accord’s use in Cassandra with simple application examples.
AI-written summary · about
chapters (14)
- 0:00Data replication intro
- 1:25State machine replication
- 3:51Commutative operations
- 5:35Leaderless limits
- 7:25Égalitarian Paxos overview
- 13:25Why Égalitarian Paxos unused
- 16:15Latency control with Accord
- 20:12Accord in Cassandra
- 25:15Accord demo
- 29:51Fault tolerance robustness
- 33:01Égalitarian Paxos spec issues
- 36:14Accord protocol validation
- 37:20Accord performance
- 40:00Q&A
related talks
34mThe Accord Consensus Protocol: ACID transactions globally at any scale
48mThe Accord Consensus Protocol, how to do ACID transactions at scale
30mApache Cassandra: What’s New and What’s Next
54mEvolving Transactions in Apache Cassandra
34mCassandra on ACID: This Changes Everything
51mDo Not Reimplement Distributed Consensus Algorithms
53mApache Cassandra. Inception - all you need to know
40mConsensus Algorithms in Distributed Systems