Content distribution on the Web is moving from an architecture where objects are placed on a single, designated server to an architecture where objects are replicated on geographically distributed servers and clients transparently access a nearby copy of an object. In this paper we study how the different redirection schemes used in modern content distribution networks affect the user-perceived performance in normal Web page viewing. Using both simulations and experiments with real Web servers we show that redirection schemes that require clients to retrieve different parts of a Web page from different servers yield sub-optional performance compared to schemes where a client accesses only one server for all the parts of a Web page. This implies that when replicating Web pages, we should treat the whole page (HTML and images) as a single entity.
ASJC Scopus subject areas
- Computer Networks and Communications