Containerized infrastructure allows workloads to share hardware via an orchestration system such as Swarm, Kubernetes, Mesos, Fleet, Nomad etc. The higher your hardware utilization, the more efficient your cluster. The key to efficient utilization and maximum performance lies in the logic that decides how to place workloads: the scheduler is what makes the difference between a mediocre and a stellar cluster!
Ionel and Malte will give an overview of the state of the art in cluster scheduling, focusing both on industry practice and bleeding-edge academic research, and will discuss why good scheduling matters to cluster owners, developers, operations staff, and end-users alike. They will describe the Firmament scheduling platform, designed to make excellent placement decisions, flexibly enforce user-defined scheduling policies, and scale to tens of thousands of machines and hundreds of thousands of containers. Finally, they'll discuss ongoing integration efforts to bring Firmament to existing orchestration systems as a pluggable scheduler.