| Today, along with booming development in network technology, Internet has become the fundamental infrastructure of an information society. It is indicated in some statistical data that HTTP-based Web traffic has accounted for 75～80% of total Internet traffic, and this proportion is still rising. However, HTTP application is based on the client/server pattern that is essentially not extendible. Therefore, enormous access to the same Web server will make the backbone network and server-located network congested, users' latency time too long, and even make servers too burdened to be accessed.Having adopted the technology such as caching, duplicating, load-balancing, advanced routing, request redirecting and so on, CDN (Content Delivery Network) of Internet is apt to distribute information resources even to the whole network, which makes Internet users accessing required contents from the "closest and best" servers feasible. This will save backbone bandwidth, reduce source-server's traffic, and finally can greatly improve access speed and service quality for terminal users. CDN is a new concept whose aim is to construct a geographically distributed content delivery network by adding a new network structure based on the existent network hierarchy. Implementation of CDN depends on various network technologies, of which technology such as caching, dynamical content distribution and duplicating, load-balancing and so on is primary. Since Web-caching is one of the basic and primary technologies, it is essential for implementation of CDN in Internet to study the technology of Web-caching, design caching system and analyze its performance. This paper has researched into the technology of CDN, the theory of Web caching and ICP(Internet Cache Protocol) — a Web-caching protocol in prevalence recently. Concerning the architecture, some implementation methods for Web-caching system design are discussed: single server high-speed caching system, distributed high-speed caching system, hierarchical high-speed caching system and high-speed caching network. In Supported by the National Natural Foundation of China under Grant No. 60103005addition, advantages and disadvantages of different implementations are also discussed. After analyzing the performance index of high-speed caching system, we bring forward and implement a new Centralized Managed Web-Caching System. Compared with ICP-based Hierarchical Caching System, it is theoretically proved that our Centralized Managed Caching System has more efficient buffer, lower process cost, less user access latency and better scalability. In Centralized Managed Caching System, prearranged algorithms dispatch users' HTTP requests to certain appropriate caching server in system, which eliminates large communication and processing cost, and finally reduce the content redundancy between different caching servers of the caching system.According to this paper's methods for researching the design of high-speed caching system, we have constructed single server architecture mode, hierarchical architecture mode and centralized managed architecture mode of CERNET (China Education and Research Network) National Central Web-Caching System using Squid (a popular web-caching software).The caching system generated log files are analyzed, and the timing automatic processing analysis and real-time monitoring functions for the system log files are implemented. By comparing the log files of Hierarchical Web-Caching System with those of Centralized Managed Web-Caching System, we validated the superiority of the Centralized Managed Web-Caching System.Centralized Managed Web-Caching System designed and implemented by this paper avoids the limitation of communication cost and processing ability among caching servers, making different caching servers share in users' requests more efficiently and reducing information redundancy in system. This Centralized Managed Web-Caching System has applied in CERNET, and its performance indices such as system shooting ratio have distinctly improved.
|
PDF Full Text Request