KEDA, HPA, and VPA? An Introduction to Scaling a Event-Driven Workloads on Kubernetes

Track: Unobtanium
Scaling modern event-driven applications that use an event-loop, such as those written using Node.js or Vert.x shouldn't be based on a single metric such as CPU usage. Code that uses blocking APIs or performs CPU-bound operations can have a significant impact on application responsiveness. Using Worker Threads can help alleviate pressure on the event-loop, but scaling out more replicas also plays an important role. Join this session for a practical introduction to Kubernetes autoscaling, and KEDA. We'll examine an application that suffers reduced throughput and increased latency due to CPU-bound tasks blocking the event-loop. We'll then explore how Worker Threads and autoscaling features provided by Kubernetes and KEDA can keep our application both highly-available and responsive.
Evan Shortiss
Evan Shortiss is a Developer Advocate at Red Hat, where he helps developers to be successful with cloud-native technologies. He has more than ten years of development experience, with a focus on building web applications and backends using JavaScript ecosystem technologies.