technicalreports

Advisor: Dr. Javed I. Khan
Department of Computer Science
Kent State University
August 2002

User perception of the responsiveness has been an important factor for Web systems. Caching and prefetching are the two principal techniques known to improve responsiveness for systems that involve data communication. While caching helps in case of repeated references, prefetching reduces response lag for new references. There are various factors that affect prefetching performance. This thesis studies how the user reading and interaction behavior and the organization of the Web documents affects the response time and amounts of transferred data in prefetching performance. In order to achieve this goal, a HTTP proxy system called RHDOS is developed to simulate the performance of prefetching. Experiments are based on four different types of Web documents organization including chain, tree, fully connected graph, and a complex tree with core graph. Three performance impact factors have been selected: user interaction duration, reading sequence, and prefetching sequence. Analysis results show that, compared to a matched system, the response time of a random system can take 1.6 - 6.3 times larger and bring in 1.8 - 2.0 times more unnecessary data. In the worst case, a completely mismatched system's response time can be about 1.7 - 11.3 times larger and result in 1.3 - 1.4 times more unnecessary data than a matched system. Smarter prefetching techniques can be developed if we take the structure of webspace and user reading behavior into consideration. This study may also help content developer organize the webspace so that it can be navigated faster.

Thesis Title:

Impact of Webspace Organization and User Interaction Behavior on a Prefetching Proxy