Design And Implementation Of Web Cluster Load Balancing System

With the rapid development of Internet, the working life of people is more and more inseparable from the Internet.By December 2013, there were 618 million Internet users in China and the Internet penetration rate was 45.8%. The rapid increase of Internet users has brought great challenges to a lot of popular websites. Some of them have to deal with millions of requests everday, the explosive growth number of visits makes establishing a stronger network service system an urging problem. To provide users with high-quality experience and to solve the problem that network traffic surge demands more for network bandwidth and server performance.The vast majority of sites have built a web server cluster to improve the processing capacity of the system. Web cluster technology is one of the key technologies in the construction of large sites.The subject of this thesis is Base web server cluster load balancing scheduling.The thesis studies some key technologies of web cluster, including the key technologies of implementing a load balancing systems and the load balancing scheduling Following results have been concluded: 1.The thesis designes a new scheduling strategy based on the content of the HTTP request.When a request arrives,according to its content,the strategy decides the type of the request:dynamic request or static request.The strategy has different scheduling decisions for different requests.For dynamic requests,its symmetrical distribution of load to consider,for static requests,the main consideration is to improve the Cache hit ratio. And fuzzy clustering dynamic feedback regulation mechanism is used to ajust the load distribution of cluster. 2.The thesis have implements a load balancing system.Specially,the design and implementation of the load balancing system is described in detail.In the system,we use epoll events drive model, hash table, connection pool as key design technologies.And the high performance of load balancing system is implemented.Finally, a test environment is built,simulate multi-user stress test on the load balancing system is conducted.It also analyzes the performance of the strategy proposed in this thesis, in comparison to LARD, and the weighted least connection algorithm.