talkTuning docker network for large scale production workload

Are you using docker containers at scale? Say on hundreds to thousands of machines? Have you hit performance issues while running thousands of containers, docker daemons and networks? If yes, this session if for you. We manage a cloud offering that runs on 1000+ machines running docker daemons. What started smooth deployment with small number of machines, became a nightmare when number of docker daemons scaled beyond 250; We encountered rapid performance degradation. Docker Network was extremely slow, containers start up time went up significantly and uptime of network metastore became a challenge. We had to spend days and night debugging, tweaking various configurations of docker daemon, network metastore and even some hardware components to get acceptable performance. In this talk we take you through the journey of how we arrived at better configurations that helped us scale running containerized workloads on 1000s of machines.