Key points are not available for this paper at this time.
The World Wide Web and the phrase 'traffic jam' have become as linked in the minds of many computer users as are the urban superhighway and rush hour to the early morning commuter. Insufficient bandwidth causing high latency are the order of the day. Caching is a standard solution for this type of problem, and it was applied to the Web early on for this reason. The more advanced approaches of replication and application level naming were introduced to further attack the problem. However, these complex methods could easily exacerbate the situation if the implementation is not well designed and the results are not evaluated carefully. This article examines the benefits that can be obtained from both caching and replication by analyzing the performance of the main proxy cache at the University of Kaiserslautern as well as a system of replicate servers designed and implemented at the University's System Software research group. Our data show that replication in particular helps to reduce overall bandwidth as well as user-perceived latency. It can also provide a more fault-tolerant and evenly balanced system.
Baentsch et al. (Wed,) studied this question.