Building APIs at scale - Caching and how to avoid cache stampedes
Presented by: Matt Nedrich
As an API scales, it will likely need caching to remain performant. Cache stampedes are a type of cascading failure that can occur in systems under heavy load, as these systems attempt to service many concurrent requests and also keep their caches up to date.
There are several approaches to mitigate cache stampedes - including external recomputation, locking, and probabilistic approaches.
At Gambyt we build software solutions for the regulated gaming industry, and manage APIs for clients that receive hundreds of millions of requests per week.
This talk will highlight some of our learnings around scaling APIs, caching, and review strategies for mitigating cache stampedes.