Design And Implementation Of QOS And Load Balancing System For Proxy Cluster

Proxy is a common used network service. Compared with traditional direct network connection, it can increase the network access speed, restrict the requests for network resources and protect user's privacy. However, proxy faces two problems. One is how to meet the transmission requirements of proxy service given that different proxy business requires different quality of service. The other is how to optimize the quality of proxy service in the network with proxy server cluster forwarding through multi-line. To solve these problems, methods should be taken to balance tasks assignment, reduce transmission delay and maximize proxy utilization.Based on the difference in proxy business Qo S(Quality of Service) requirements and some applications fail to use proxy. This paper, at first, forwarding a method using custom protocol chain to transmit application data, evaluating and selecting proxy server based on the business types. The protocol chain consisting of layered service provider and base service provider, are used in traffic transporting which is classified by application names. Then we put forward an algorithm for proxy evaluation. Firstly, set weights for Qo S indexes according to their importance for proxy business. Secondly, evaluate alternative servers employing attribute optimization method and ideal solution method to find the optimal proxy server that best suit the proxy business.Aiming at balancing load of servers in the proxy server cluster, experimental test and analysis are carried out on scheduling effect of static polling algorithm and dynamic Weighted Least-Connection Scheduling algorithm. To overcome shortcomings of the above two algorithms, this paper provides a load balancing algorithm based on the theory of maximizing deviation. The algorithm is briefly described as follow. Firstly, collect proxy server's load status and network status. Secondly, dynamically adjust attribute weights according to deviation between the attributes and then generate comprehensive evaluation for each node. Thirdly, take the radio of the current traffic amount to the comprehensive evaluation as the final weight for the node. Then, assign tasks to each node according to its weight in one scheduling cycle using Weighted Round Robin method. The experimental results show that the improved algorithm can effectively reduce the average response time applied to the heterogeneous proxy, and improve the utilization rate of proxy cluster.Experimental test and analysis are carried out on the optimal proxy service evaluation algorithm and the improved load balancing algorithm. Then a proxy system is designed and implemented to validate the performance of the methods. The experimental results show that the designed proxy system can balancing proxy server load according to their performance and state, improve the network access speed, reduce 53.9% of the response time while user accessing the network, and the proxy system also has a lower server resource utilization.