Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore load management and distribution challenges in Apache Cassandra during this 48-minute technical talk presented by Alex Petrov. Dive into the Staged Event-Driven Architecture (SEDA) threading model implementation and understand why finite queue sizes were never implemented and how request processing lacks deadline awareness. Learn about metastable failures, the implications of hidden latency metrics, and debugging techniques for failures and livelocks. Discover how effective load management requires collaboration between clients, coordinators, and replicas, with nodes explicitly communicating their capacity limits and actively managing excess load. Understand how clients should respond dynamically to overload signals and propagate them through the stack. Gain insights into intermediate mitigation strategies and future directions for Apache Cassandra's load distribution, queue sizing, timeout handling, and intelligent routing mechanisms.
