Design And Implementaion Of The Distributed System Resource Management And Deployment Platform

With the development of Internet and the arrival of the mobile Internet era,in order to cope with the business demands of bigdata,the scale of the cluster is increasing,which leads that the problems of automated deployment and management for these distributed systems become more prominent.While there are already automated deployment tools such as Ansible,which are able to handle some of deployment and management tasks.But fully automated deployment and management also continue to face significant challenges.The distributed system resource management and deployment platform to be described in this paper can provide automated deployment and management services.Instead of separating service development and deployment management,the system combines them by taking advantage of a component layer.To increase the scalability,this system allows users to customize the resource statistics strategies and allocation strategies.The system can meet the different resources requirements of the service and perform reasonable allocation of nodes.The platform not only provides lightweight interfaces for system resource management but also has the feature of cluster's automation deployment and management.The system which relies on COM-based component framework OCF,the underlying network communication engine Io Engine and the RPC-like message communication component Channel consists of three subsystems:1.The underlying distributed system resource management subsystem,whichorganizes the nodes in the cluster via P2 P network,provides a unified,scalableinterface for accessing resource usage of the current node.2.There is also a basic subsystem,called the component pool subsystem,fororganizing all the components.It allows users to register,update,delete,andobtain the components in the component pool.3.The distributed deployment platform,which provides the ability to automatethe deployment and management of distributed clusters,is built on distributedsystem resource management subsystem and component pool subsystem.Itcan automatically allocate nodes and achieve load-balancing through theinterface of the underlying system resource management subsystem.Meanwhile,different users can use the interfaces of the platform to customize cluster management scheme according to the demand by themselves,which shows the good scalability of the system.This thesis will introduce the related technology and theoretical basis firstly,then focus on requirements analysis,design as well as implementation,and finally show the results of function test and performance benchmark.The system is developed by C++ under the Free BSD and relies on OCF,a COM-based component framework.The test results show that the component-based asynchronous programming model helps the system to achieve a good performance.