Design And Implementation Of Scalable Architecture Based On Server Cluster

In the gradual evolution process of the website,by changing the number of deployed servers rather than changing the design of the website's hardware or software,the goal of expanding or reducing the service processing capability of the website is archieved.This is the scalability of the website.The most important way to achieve website scalability is taking server clustering technology.This thesis studies the evolution process of internet site architecture and the related technologies of server clusters.Based on Open System Interconnection Reference Model(OSI/RM),it summarizes the principle,advantages and disadvantages of various clusters and load balancing techonology,and determines the design ideas of mixed cluster of Linux Virtual Server(LVS)and the reverse proxy(Nginx).Through the study of the infrastructure of the major websites,it also determines the feasibility and necessity of the design of the website's scalability structure.This paper analyzes the implementation principles and functional characteristics of traditional high availability clustering solutions.It proposes the design concept of Multi-path Access(MA)network cluster inview of the shortage of the design of one main standby and standby machine idle.The cluster of three-layer switch and LVS is built with the Open Shortest Path First(OSPF)link state check principle,by configuring dynamic routing on the Layer 3 switch,and the route simulation software(Quagga)deployed on the LVS.The request will be forward to Nginx by the LVS direct route(DR)model configuration.The scalable design site architecture is built with a hybrid cluster of multiple LVS collaboration clusters and Nginx reverse proxy clusters.The improved scheme has the following advantages.First,high availability cluststers do not have resources to idle and avoid wasting any resources.Second,the true high availability is archieved,the failover of the whole station service will not take place.Third,it breaks through the restrictions of up to two servers in the traditional high availability scheme,supports more service clusters,and enables the websites to have larger capacity.The hardware configuration of the experimental environment in the paper is scaled down,and test cases are designed around the core functions of the solution.And the validaty of the reverse proxy,the health checks back-end server,the validity of LVS clusters,the load balance during a large number of requests,the functional performance of high-availability solutions and all verified.In addition the necessary performance tests,stress tests and stability tests are carried out in the experiment environment.The experiment initially verified the feasibility and effectiveness of the paper's design.The design of this article has been promoted in the production environment of the Open Class,and the scalable architecture has supported in the twice previous peak of site visiting.At the peak of 2016,the one day page views(PV)reached 50 million,and the number of unique visitors(UV)exceeded 1 million.At the peak of 2017,PV exceeded 70 million(PV 500,000 and UV 80,000 in common days).The production environment is designed and transformed with the hybrid cluster architecture of LVS,Nginx,OSPF,coupled with the caching technology(CDN,Session Convergence),to achieve high-availability operations without disruption during peak hours and without major failures on the site.It has proved that the design of the website architecture of this article is clear and the risk of implementation is low.It can meet the requirements of high concurrent access and long-term operation of the website during the site's site's peak period.