Handling 80 Petabytes A Month: Autoscaling with Traffic Volume to Protect Revenue and Reputation
Every second you experience downtime a customer is unhappy, revenue could be lost, and your product’s reputation is at risk. HAProxy, the fastest load balancing software in the industry, has been leveraged by Docker to build a cloud native API gateway system that is scalable, resilient, and highly available. In this presentation, Ryan will demonstrate how to leverage Kubernetes (using Amazon EKS) node and pod autoscaling to respond to traffic volume and will cover configuration tips to maintain a highly performant HAProxy environment. This approach has allowed Docker Hub to scale upwards of 35k requests/s, transfer 80PB/mo and avoid downtime for all deployment, configuration, and traffic requirements. This conversation is for software architects who are interested in building scalable, resilient, and highly available cloud native HAProxy deployments.
Here you can view the slides used in this presentation if you’d like a quick overview of what was shown during the talk.
Infrastructure Engineer, Docker
Ryan is an engineer at Docker, working on the Cloud Platform team responsible for building, designing, and maintaining the Docker Hub infrastructure. He has a passion for highly scalable cloud native systems, focusing in particular on Kubernetes and service meshes. In his spare time he enjoys skiing, live music and reading fantasy novels.