Skip to content

Latest commit

 

History

History
51 lines (32 loc) · 3.35 KB

README.md

File metadata and controls

51 lines (32 loc) · 3.35 KB
Raw

Edg-cloud System

Introduction

My zero-copy undergraduate work (outstanding final design awarded), a K8s-based edge cloud system consisting of (i) a front-end based on Rancher and Echarts visualisation with lots of interesting interactive animations, and (ii) multiple back-ends supported by classical, intelligent scheduling algorithms.

Video demonstration

https://www.bilibili.com/video/BV1ZX4y1o7vh/


Project Overview

The cloud edge system architecture is as follows:


  • Device Side：The device side is a workload generation. In order to simulate the generation of user requests, the request generation is simulated by using Ali Cloud and other environments as request generators.
  • Edge Side：Requests received by the edge-side entry access point (eAP) are forwarded by the scheduling algorithm to the edge cluster or cloud cluster for request processing.
  • Cloud Side：Compared to the edge cluster, the cloud cluster has more powerful computing and processing capabilities, and more complex and intelligent applications are deployed in the cloud.
  • Private Registry：A private pegistry is configured for the system platform, in which a large number of service images are stored to facilitate the deployment of services in each cluster.
  • System Monitor：System information such as the number of nodes, resource utilisation, real-time scheduling of tasks, processing of tasks and deployment of containerised services, etc., will all be presented by the front-end.

Part of Implement and Design

PVE virtual cluster construction:


Modification and recompilation of Rancher (a K8S monitoring open source architecture)：


Custom visual web creation based on the Echarts library. In fact, I am very proud of the implementation of the central topology diagram. Because, using the current visualisation library to present the topology is extremely easy to be limited in terms of interactivity. For the topology of the system, I have combined the various drawing modes of the Echarts library to ensure a high level of interactivity：


Containerisation and deployment of several popular GitHub AI applications：


The system back-end contains greedy algorithms for container services and deep reinforcement learning scheduling algorithms：