PaaS Application Management Platform Design And Implementation Based On Kubernetes

Cloudification is the main theme of the IT industry in recent years,Cloud-Native and microservice architecture is in the ascendant,new cloud technologies such as container and PaaS platform originated from Internet companies,but also have board application space in traditional industries such as telecommunications,power and oil,this thesis proposes a PaaS application management platform based on Kubernetes with the characteristics of applications in the telecommunications industry.The main work of this thesis is as follows: 1)Focus on how to build a PaaS application management platform based on Kubernetes,an overall implementation is proposed,and analyzed under a typical enterprise application scenario(Business and Operation Support System),two alternatives are provided(service gateway and configuration management)as the original components does not fulfill the requirements,so that this solution can support the complex architecture and management requirements of the enterprises.2)This thesis optimizes the resource scheduling function of the Kubernetes,proposes a prediction-based optimization.The original scheduling algorithm of Kubernetes only considers the resource requirements of a certain time,which is a static value.This thesis introduces a machine learning algorithm(Prophet)to learn the characteristic from the historical data,and the characteristic is applied to the priority calculation of the scheduling algorithm to achieve more accurate scheduling.The optimized scheduler is compatible with the original scheduling mechanism,so it has better practicality.3)Experimental verification tests were performed on the work of this thesis.The platform that based on this thesis has been launched in a dozen of projects overall the world,has been proven to be efficient.