Cluster In The High-speed Network Design, Web Caching System

With the rapid expansion of Internet, the improvement of network bandwidth can not catch up with the increasing of network users. Nowadays, the increasing applications of web service lead to a conflict between the limited bandwidth and user experience. Therefore, in order to reduce the network outlet traffic and enhance the user experience, it is necessary to research and develop a cluster web cache system.As a sub-system of"P2P traffic cache system", researched by Institute of Acoustics, Chinese Academy of Sciences, High-performance network laboratory, this new type of cluster web cache system is deployed in the main nodes of internet to carry out application-layer protocol analysis and flow control. At present, it has been put into use by more than 20 domestic telecom operators.At the beginning of this thesis, series of application and current research home and aboard related to the cluster web cache system are introduced. This thesis makes a detailed discussion of HTTP protocol related to web cache system through expounding the web cache system theoretical foundation. Then, it also makes a detailed analysis and comparison about the different technologies with respect to the problem of high-speed network packet capture and load balancing. In the rest chapters of this thesis, it describes the basic function of packet capture module, load balancing module, and linker analysis module. The results of tests show the design of cluster web cache system have obtained expected scalability and I/O performance, also some drawbacks of its implementation are mentioned in last chapter.The major contributions of this thesis are to solve seven problems as follows:1).Proposing a kind of transparent, front-end and back-end of the loosely coupled, high-performance cluster web cache architecture, which is suitable for being deployed in high-speed network.2).Pattern-oriented software architecture. Through importing the form of components, cluster web cache system can add HTTP module to expend the original function of P2P traffic cache system.3).Using zero-copy technology to capture HTTP packets in high-speed network, through the transport layer filtering to reduce the workload of application layer filtering.4).Load balancing scheduling of cluster web cache system is based on the document feature value.5).Using the technology of picking up document feature value to merge the web content.6).The development model of web server is based on event-driven mechanisms.7).To design and achieve high performance, hash function, memory pool and thread pool are used to speed up searching and transmitting files.