Seamlessly and Coherently Locating Interesting Mirrors on the Web

Neilze Dorta, Pierre Sens, Christian Khoury

Nowadays, the World Wide Web is used mostly as a common medium for information sharing. Therefore, locating an object on this large scale dynamic medium tends to be more and more difficult. Content Distribution Networks, e.g. Akamai, and global naming services, e.g. Globe, do more or less than what is required by most users. In this paper, we are interested in discovering, advertising, and transparently locating interesting mirrors of interest to a group of users. Our solution, AR , is user-centric; it uses cooperation among organizations to discover, publicize and locate coherently new mirrors that are of interest to them. Access transparency is achieved through a naming service that manages the different aliases for the same replica. Consistency guarantees are given to each user that no document delivered would be older than the one viewed before. The system scales geographically due to the epidemic and asynchronous nature of the cooperation protocol. We propose a methodology for creating homogeneous groups with common interests, using collected Web traces, then give a glimpse of the potential benefits made by using AR . It opens a path towards making mirroring ubiquitous, hence fostering a better use of the Internet and its resources. A prototype has been implemented in Java and will be used, in the future, in real-world tests for more accurate and realistic results.

Full Text