Скачать или смотреть Efficient Resource Utilization for Batch Compute on Kubernetes - Amit Kumar & Kevin Xu, Uber

Don't miss out! Join us at our next Flagship Conference: KubeCon + CloudNativeCon Europe in Paris from March 19-22, 2024. Connect with our current graduated, incubating, and sandbox projects as the community gathers to further the education and advancement of cloud native computing. Learn more at https://kubecon.io

Efficient Resource Utilization for Batch Compute on Kubernetes - Amit Kumar & Kevin Xu, Uber

At Uber, we run multiple types of batch workloads on Kubernetes at large scale (tens of thousands of workloads running concurrently on millions of cores). Workload owners are unaware of resource requirements from each other which may result into in-efficient resource utilization overall. The Compute Platform team solves the problem of efficient utilization of resources. Scale limit of the Kubernetes scheduler controls the size of a Kubernetes cluster, so we run workloads on a multi-cluster setup. Efficient utilization of resources becomes even more challenging due to the problems like fragmentation and hot-clusters in a multi-cluster setup. This session focuses on challenges and solutions of running batch workloads in a multi-cluster setup. We will talk about Federation of batch workloads which solves problems like fragmentation, uniform load sharing across clusters, resource sharing across pools, unified view of available resources and the metrics governing the overall efficiency.