Autothrottle: A Practical Bi-Level Approach to Resource Management for SLO-Targeted Microservices

Explore a conference talk from NSDI '24 that introduces Autothrottle, a bi-level resource management framework for microservices with latency Service Level Objectives (SLOs). Delve into the challenges of achieving resource efficiency while preserving end-user experience in cloud applications adopting microservices. Learn how Autothrottle architecturally decouples application SLO feedback from service resource control, bridging them through performance targets. Discover the implementation of an application-wide learning-based controller that periodically sets performance targets for per-service heuristic controllers. Examine the evaluation results on three microservice applications using production workload traces, showcasing significant CPU savings compared to baseline approaches. Gain insights into this award-winning research that addresses the complexities of resource management in modern cloud architectures.