java.util.concurrent for Distributed Coordination
summary
Hazelcast implements linearizable distributed java.util.concurrent primitives on top of the Raft algorithm: AtomicLong, AtomicReference, CountDownLatch, Semaphore and a FencedLock that returns a monotonic fencing token via lockAndGetFence. Sessions with TTL track clients and auto-release resources; fencing tokens prevent stale-owner actions and resolve TTL trade-offs. Features include CP groups, dynamic cluster membership, Jepsen-style testing, demos showing an AtomicReference+CountDownLatch config flow and FencedLock leader election, and ongoing work on persistence, a KV API, events and management tooling.
AI-written summary · about
chapters (9)
- 0:00Introduction and agenda
- 4:36Distributed coordination challenges
- 10:58JDK concurrency for distributed systems
- 17:44Raft consensus algorithm
- 26:34Demo: Configuration management
- 30:50Demo: Leader election with FenceLock
- 37:10Demo: Fencing tokens for safety
- 40:25Summary and future work
- 43:21Q&A and giveaways
related talks
47mjava.util.concurrent for Distributed Coordination
41mWhat can go wrong with the distributed system - the Hazelcast exp...
48mHow to build distributed concurrency primitives
46mWhat can go wrong in a distributed system - experience from...
30mDistributed Computing with Hazelcast
55mHazelcast - Hitchhiker’s Guide. Grzegorz Piwowarek. Java Fest 2020 Online
59mDistributed applications using Hazelcast
44mJava: Introduction to the Hazelcast Ecosystem