next up previous
Next: Searching through a Up: Problem Description and Previous: Problem Description and

The Cause and A Solution

The major cause of the problems discussed previously is that search and index services are too centralised. Whilst the Web documents are distributed and scattered at different places on the Web, databases services are centralised with their own servers. An ideal solution is to have a distributed and local search process.

A distributed and local search process would mean the searching is done at many servers around the Web, and not just at one or two servers. This is like having a database service at every server. This would solve the problem of accessing. And as the number of Web users increase, the accessing problem will not worsen because the number of Web sites would increase proportionally.

figure=ideal.eps Ideal Database System

figure=actual.eps Feasible Database System

Although the above solution is logical, it is not feasible. It is impossible to have an index of all documents on the Web contained in every Web server. A more feasible solution is to have :

Each of the indexes in the Referring List should be small in size, and only contain necessary information, so a large number of these indexes can be stored in a Referring List at each server. So how could the searching, updating, and making be done in this system?





Tommy Wing Yiu Tsui
Tue Nov 7 10:21:32 EST 1995