Distributed Web Crawler

Abstract

This thesis investigates possible improvements in distributed web-crawlers. Web-crawling is the cornerstone of search-engines and a well defined part of Internet technology. Due to the size of the Web, it is important that a web-crawler is fast and efficient, since a web-crawler should be able to find the interesting sites before they change or disappear. The thesis will focus on crawler distribution concerning modularity, fault-tolerance and group membership services. The download order of crawlers will also be covered, since this greatly influences the efficiency of a crawler. In addition to the theoretical basis of the thesis, a prototype has been constructed in Java. The prototype is efficient, modular, fault-tolerant and configurable. The result from the thesis indicates that using a membership service is a good way to distribute a crawler and conclusively, the thesis also demonstrate a way to improve the crawling order compared to a breadth-first ordering.