Detecting and Addressing Overload in LinkedIn Microservices

Explore the innovative Hodor (Holistic Overload Detection & Overload Remediation) framework developed by LinkedIn in this 44-minute conference talk from Strange Loop. Discover how this system detects service overloads from multiple potential root causes and automatically improves the situation by dropping just enough traffic to allow services to recover. Learn how Hodor maintains optimal traffic levels to prevent services from reentering overload, all without manual tuning or threshold specification. Gain insights into the framework's overview, overload detection methods, and request dropping techniques for providing relief. Benefit from the expertise of Bryan Barkley, a Sr. Staff Engineer at LinkedIn with over two decades of experience in server-side Java, as he shares his knowledge on improving system resiliency and availability in microservices architecture.