Campus Network Server Cluster Based On The Linux Load Balancing And Reliability Technology

The human society has been entering a network information era, in which users need faster, more reliable and functional network services. With the rapid network construction, more and more Area LAN, such as Campus Network. Intranet. Metropolitan Area Network, were built. But the capability of server on the Area LAN can not afford the increasing demand of growing users and network traffic. So server cluster is used to enhance the capability of Area LAN. The load balance addressed in this thesis is the key technology in server cluster.First, on analysis of general structure of Area LAN, we point out three potential bottleneck of performance: user access server, Proxy server with cache, and application server (such as Web). The technologies of Load Balance need to be applied. In addition, the Single Point of Failure still existed in some key nodes, so some methods must be used to enhance the reliability of system.Second, we analyze and evaluate many current technologies of Load Balance. Their principles of implementation are also elucidated. And based on the analysis of their architecture, we present two architectures: concentrated and distributed architecture.At last, according to the environment of Area LAN, we present a total solution based on Linux. These solutions include 4 parts: · The design and implementation of PPPOE Access Concentrator. Current PPPOE protocol can not be used to implement an Access Concentrator Cluster for without the function of load balance. So we add the function of load balance to this protocol. · The design and implementation of Internet Cache Dispatcher (ICD) system. The Proxy server with cache is commonly used in current Area LAN, when users rapidly increase, it will be a bottleneck. So we design and implement the ICD system. It can dispatch the users' request packets to different Proxy server according to their destination address. So it can balance the load of Proxy servers, and increase the Cache Hit Rate. · The analysis and improvement of Linux Virtual Server (LVS) system. LVS system can be used to construct an application server cluster when one single application server (such as Web) can not afford the load. But there exist the problem of user connections losing, so we present a method of user connections' dynamic backup. · The analysis and improvement of a Standby system based on Heartbeat. Heartbeat software can be used to construct a Standby system to resolve the problem of Single Point of Failure in key nodes of Area LAN. But there are still some veracity and agility problems in failure check. We improve it so that it can work in various applications in Area LAN.These 4 parts above can be used separately or work together in the Area LAN. So we can get an Area LAN of highly performance, highly scalable and highly available. This thesis discusses the detail on how to design and implement the total solution.