Research And Implementation Of Dynamic Equilibrium Algorithm For Server Cluster Based On Tengine

With the continuous improvement of mobile network technology,applications are gradually connected with people's daily life,such as Wechat,Taobao,etc.However,because of the widespread popularity of Internet applications,the number of connected access of the back-end server is booming,which aggravates the load on the back-end servers.In order to solve this problem effectively,the thesis researched and implemented a high-performance server cluster system.We researched the load balancing technology and distributed database technology of front-end and rear-end access in high-performance server cluster system.First,For high concurrent access of Web server,we used asynchronous I/O multiplexing decoupling back-end business layer instead of the FastCgi synchronous access mechanism.In order to solve the problem of C10 K concurrent connection,this thesis designed a new dynamic weight algorithm of back-end server which based on Tengine server front-end load balancing principle.Through self-compiling Ngx_Lua scripting language module and Redis memory database module,it collects server load evaluation standard and implements custom dynamic weight load balancing,and enhances the Web layer dynamic access performance.Second,when dealing with the data in the backend server business layer,we deploy a distributed database cluster system,based on MySQL database binary log,to achieve multithreaded synchronization information collection,to complete master-slave synchronous replication of distributed database.For the feature of this thesis IP-CAMERA business project is read more and write less,we use database middleware MyCat build database cluster read-write separation,effectively alleviate I/O congestion caused by write latency,and steadily improve data throughput.Finally,By the HTTP server cluster test tool AutoBenth to simulate the real client C10 K concurrent access,we did a lot of experimental tests and data acquisition in the test environment which was built on Ali cloud server.Through comparing and analyzing the data collected under different load balancing algorithms,the custom load balancing algorithm implemented in this thesis enhances the C10K's concurrent access performance effectively,reduces the data's error rate,while the back-end distributed database cluster running stable,and the overall server cluster performance has improved.