In this project, we will talk about Web caching systems. In fact web caching is generally considered an efficient way to enhance World wide web overall performance making it scalable. In this report, we take a look at the design and evaluation of Web caching systems. To begin with, we recommend a new basis for making caching choices; specifically, a site-based technique. In contrast to the focus of traditional methods, the importance of this technique is to take into account only the Web site which an object is associated with, not the object itself. We reveal that this technique provides quite a few advantages for several various kinds of applications in Web caching areas. In line with the idea of the site-based technique, we suggest a site-based mapping scheme for parallel proxy servers. The fundamental thought is to map all Web objects coming from the same website to the same proxy server. By taking benefit of the persistent connection feature offered by HTTP/1.1, this mapping scheme minimizes 40%-70% of the connection setups and teardowns in comparison with a conventional URL-based mapping scheme.
To get rid of the challenge of overloaded proxy servers, we recommend a site-based dispatching technique which, by avoiding needless requests from reaching the proxy, decreases the amount of work. Analytical and simulation results reveal that the offered approach decreased the proxy load by up to 48% and therefore allowed a system throughput gain of up to 55%. To deal with the issue of restricted storage capacity, we research the startup behaviors of caching systems utilizing the least-recently-used (LRU) cache replacement policy. We present precise analyses of cache occupancy and cache miss probability. These types of steps offer information about know how long a cache takes to become completely occupied, and just how rapidly a cache reaches its acceptable cache miss probability. These analyses are proven to be really precise by computer simulations. Last but not least, we talk about polling models which can be utilized to model numerous areas in Web caching systems. We offer a new approximation of mean-waiting time for asymmetric polling systems with nonexhaustive service discipline. Numerical outcomes reveal that the suggested approach provides a much more accurate and dependable result compared to those of another well-known approximation.
Download URL 2: Visit Now
Source: City University of Hong Kong