Content Routing in a Network of WAIS Servers

Locating and accessing information in a large distributed system is a di cult problem of growing importance. This paper reports on our experience building and using a prototype system for transparent, user-guided associative access to the contents of a large, distributed set of WAIS servers. Our system is based on content routing, an architecture that makes use of content labels for locating and accessing information in large distributed systems [12]. Our content router for WAIS servers is implemented as a Semantic File System that constructs content labels from WAIS source and catalog les. The content router guides locating documents by suggesting terms that frequently appear with a given query term in document headlines. Su ciently narrowed queries are routed to WAIS servers and processed in parallel. We have successfully used our content router to locate documents on a large number of WAIS servers. Along with demonstrating the feasibility of distributed nding in a large network of information servers, the prototype shows interactive performance limited primarily by the processing time at WAIS servers.