How Alibaba extends K8s scheduler to support AI and big data workloads

CNCF Member Online program
Presented by: Alibaba Cloud

Recorded: Wednesday July 15, 2020

Download Slides

Program Speakers: Zhang Kai, Staff Engineer @Alibaba & Wang Qingcan, Senior Engineer @Alibaba

This webinar will be delivered in Chinese.

We will introduce the development of Kubernetes Scheduling Framework and then share the best practices of supporting for complex workloads such as AI/ML and heterogeneous resources such as GPU. It also introduces how to implement job level scheduling policies, e.g. Coscheduling/Gang scheduling and Capacity scheduling.

In this webinar, you will learn:
1. Understand the latest progress and future development direction of Kubernetes scheduler
2. Learn the best practice of Alibaba cloud extended scheduler supporting job level scheduling
3. Try to build your own Scheduling Framework plugin

演讲简介:本次网研会将以中文进行, 我们将介绍 Kubernetes Scheduling Framework 的发展现状,以及阿里云 Kubernetes 服务支持调度AI、大数据等复杂工作负载和GPU等异构计算资源的实践经验。还将详细介绍如何实现Coscheduling/Gang Scheduling、Capacity Scheduling 等任务级调度特性。

1. 了解 Kubernetes Scheduling Framework 当前进展与未来发展方向
2. 了解阿里云扩展 Scheduling Framework 支持任务级调度的开发经验
3. 尝试构建自己的 Scheduling Framework 插件


Related upcoming programs