Explore a conference talk from USENIX ATC '23 that introduces Sponge, an innovative stream processing system designed for fast reactive scaling using serverless frameworks. Learn how Sponge addresses the challenges of unpredictable and rapidly changing streaming workloads by efficiently absorbing sudden increases in input loads. Discover the system's unique approach to tracking metrics, making quick scaling decisions, and implementing optimization logic at compile-time. Understand how Sponge leverages serverless framework instances to reduce tail latencies and operational costs compared to traditional VM-based methods. Gain insights into the system's performance improvements, including an 88% reduction in 99th-percentile tail latencies and an 83% cost reduction when handling unpredictable bursty loads.
Overview
Syllabus
USENIX ATC '23 - Sponge: Fast Reactive Scaling for Stream Processing with Serverless Frameworks
Taught by
USENIX