Research And Implementation Of Network Load Balance On Multi-core System

With the rapid development of Internet in business and the continuous enhancement of network-based data traffic growth, the traditional stand-alone services has become a bottleneck to meet the needs of practical applications; and network load balancing is to achieve multiple servers working together and parallel processing means,which can greatly improve server performance, and make full use of network resources. After analyzing the characteristics and advantages of load balance technology, under the background described above, we achieve the realization of the 4th layer load balance on multi-core firewall products and COMWARE software platform of H3C.This paper introduces the designment and implementation of the 4^th Layer network load balancing technology in multi-core firewall. After introducing the LB system framework, the paper gives a further study on the Load Balance including the LB principle,mechanism,function,packet forwarding strategy,scheduling algorithm,health check and persistence method. In the realization phase of design, there are the general design,demand design and detail design procedures with the golden rule of software engineering for the goal of practicability. The paper describes of the test environment, test examples,test procedure and simple analysis of test results is given.The features in this paper are: Against the particularity of multi-core system,we divide our system into control plane and data plane with combining the multi-core system development standards of H3C.The former completes all the configuration,and the latter completes all the related work with forwarding. This designment makes the program become a more real-time and high efficiency system. In the development of the scheduling algorithm,in order to complete the simultaneous visit of all the cores ,we use the design idea of every core schedule independently,which can make the system more high-powered than the traditional concept of locking on the critical resources.Futhermore,In the traditional designment of least-connections scheduling algorithm,when select the real server that has the least active connections,the new server may be the impact of large flows because of the algorithm itself.Therefore,we introduced the TCP slow start algorithm concept into our algorithm to protect the fresh server,and finally achieved the optimization program.