Partitions for Everyone

Explore the challenges and complexities of distributed systems in this 43-minute conference talk from Strange Loop 2013. Dive into the world of distributed databases, queues, and lock services, examining their durability, availability, and consistency guarantees under network partitions. Learn about common assumptions regarding system clocks and network partitions, and discover how tracking causality and reconciling divergent states impact system behavior under unstable conditions. Follow Kyle Kingsbury's six-month journey of subjecting popular distributed systems to various network partitions while under load, uncovering design limits and bugs that highlight the difficulties in building reliable distributed services. Gain insights into Zookeeper, Kafka replication, Cassandra, isolation, clocks, liveness guarantees, and testing methodologies for distributed systems.