Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/AliyunContainerService/gpushare-scheduler-extender

GPU Sharing Scheduler for Kubernetes Cluster
https://github.com/AliyunContainerService/gpushare-scheduler-extender

Last synced: 10 days ago
JSON representation

GPU Sharing Scheduler for Kubernetes Cluster

Host: GitHub
URL: https://github.com/AliyunContainerService/gpushare-scheduler-extender
Owner: AliyunContainerService
License: apache-2.0
Created: 2019-02-18T03:44:41.000Z (over 5 years ago)
Default Branch: master
Last Pushed: 2023-12-29T07:04:43.000Z (11 months ago)
Last Synced: 2024-10-29T17:55:09.529Z (15 days ago)
Language: Go
Homepage:
Size: 5.52 MB
Stars: 1,404
Watchers: 39
Forks: 308
Open Issues: 107
Metadata Files:
- Readme: README.md
- License: LICENSE

Awesome Lists containing this project

README

# GPU Sharing Scheduler Extender in Kubernetes

[![CircleCI](https://circleci.com/gh/AliyunContainerService/gpushare-scheduler-extender.svg?style=svg)](https://circleci.com/gh/AliyunContainerService/gpushare-scheduler-extender)
[![Build Status](https://travis-ci.org/AliyunContainerService/gpushare-scheduler-extender.svg?branch=master)](https://travis-ci.org/AliyunContainerService/gpushare-scheduler-extender)
[![Go Report Card](https://goreportcard.com/badge/github.com/AliyunContainerService/gpushare-scheduler-extender)](https://goreportcard.com/report/github.com/AliyunContainerService/gpushare-scheduler-extender)

## Overview

More and more data scientists run their Nvidia GPU based inference tasks on Kubernetes. Some of these tasks can be run on the same Nvidia GPU device to increase GPU utilization. So one important challenge is how to share GPUs between the pods. The community is also very interested in this [topic](https://github.com/kubernetes/kubernetes/issues/52757).

Now there is a GPU sharing solution on native Kubernetes: it is based on scheduler extenders and device plugin mechanism, so you can reuse this solution easily in your own Kubernetes.

## Prerequisites

- Kubernetes 1.11+
- golang 1.19+
- NVIDIA drivers ~= 361.93
- Nvidia-docker version > 2.0 (see how to [install](https://github.com/NVIDIA/nvidia-docker) and it's [prerequisites](https://github.com/nvidia/nvidia-docker/wiki/Installation-$version-2.0$#prerequisites))
- Docker configured with Nvidia as the [default runtime](https://github.com/NVIDIA/nvidia-docker/wiki/Advanced-topics#default-runtime).

## Design

For more details about the design of this project, please read this [Design document](docs/designs/designs.md).

## Setup

You can follow this [Installation Guide](docs/install.md). If you are using [Alibaba Cloud Kubernetes](https://cn.aliyun.com/product/kubernetes), please follow this [doc](deployer/README.md) to install with Helm Charts.

## User Guide

You can check this [User Guide](docs/userguide.md).

## Developing

### Scheduler Extender

```bash
git clone https://github.com/AliyunContainerService/gpushare-scheduler-extender.git && cd gpushare-scheduler-extender
make build-image
```

### Device Plugin

```bash
git clone https://github.com/AliyunContainerService/gpushare-device-plugin.git && cd gpushare-device-plugin
docker build -t cheyang/gpushare-device-plugin .
```

### Kubectl Extension

- golang > 1.10

```bash
mkdir -p $GOPATH/src/github.com/AliyunContainerService
cd $GOPATH/src/github.com/AliyunContainerService
git clone https://github.com/AliyunContainerService/gpushare-device-plugin.git
cd gpushare-device-plugin
go build -o $GOPATH/bin/kubectl-inspect-gpushare-v2 cmd/inspect/*.go
```

## Demo

### - Demo 1: Deploy multiple GPU Shared Pods and schedule them on the same GPU device in binpack way

[![](demo1.jpg)](http://cloud.video.taobao.com//play/u/2987821887/p/2/e/6/t/1/214292079721.mp4)

### - Demo 2: Avoid GPU memory requests that fit at the node level, but not at the GPU device level

[![](demo2.jpg)](http://cloud.video.taobao.com//play/u/2987821887/p/2/e/6/t/1/214235285109.mp4)

## Related Project

- [gpushare device plugin](https://github.com/AliyunContainerService/gpushare-device-plugin.git)

## Roadmap

- Integrate Nvidia MPS as the option for isolation
- Automated Deployment for the Kubernetes cluster which is deployed by kubeadm
- Scheduler Extener High Availablity
- Generic Solution for GPU, RDMA and other devices

## Adopters

If you are intrested in GPUShare and would like to share your experiences with others, you are warmly welcome to add your information on [ADOPTERS.md](docs/ADOPTERS.md) page. We will continuousely discuss new requirements and feature design with you in advance.

## Acknowledgments

- GPU sharing solution is based on [Nvidia Docker2](https://github.com/NVIDIA/nvidia-docker), and their [gpu sharing design](https://docs.google.com/document/d/1ZgKH_K4SEfdiE_OfxQ836s4yQWxZfSjS288Tq9YIWCA/edit#heading=h.r88v2xgacqr) is our reference. The Nvidia Community is very supportive and We are very grateful.