Distributed redirection for the World-Wide Web (extended version) Aline Baggio and Maarten van Steen October 2004 Technical report IR-CS-009 Abstract Replication in the World-Wide Web covers a wide range of techniques. Often, the redirec- tion of a client browser towards a given replica of a Web page is performed after the client’s request has reached the Web server storing the requested page. As an alternative, we propose to perform the redirection as close to the client as possible in a fully distributed and trans- parent manner. Distributed redirection ensures that we find a replica wherever it is stored and that the closest possible replica is always found first. By exploiting locality, we can keep latency low. 1 Introduction Replication in the World-Wide Web encompasses a wide range of techniques from proxy caches to mirrors and Content Distribution Networks (CDNs). One goal of these replication mechanisms is to allow clients to use the replicas that best suit their needs in terms of network distance, consistency or security. Nevertheless, the use of location-dependent URLs in today’s World-Wide Web does not facilitate transparent access to the replicated Web pages. Instead, it is often necessary to explicitly redirect clients towards a given replica. Redirection in the case of proxy caches occurs in an implicit way: each HTTP request is routed through the cache or the hierarchy of caches and, in the best case, the replica of the requested Web page is retrieved directly from the cache storage space. In the case of mirrors or CDNs, the client browser has to be explicitly redirected to a host that is normally not on the route followed by the request. Redirection is, in most cases, achieved in a home-based way. A client is redirected only after its request has reached the home server, that is to say, the host named in the document’s URL. The decision where to redirect a client to is therefore centralized. One important disadvantage of centralized redirection mechanisms is the induced latency. An- other is that the home server may become overloaded. Ideally, a client request should not be forced to go all the way to the home site in order to be redirected to a close-by replica. On the contrary, the redirection should take place as soon and as close to the client as possible. We have devised a distributed redirection scheme in which the redirection decision can be taken locally at the client machine or, in the worst case, before the HTTP request leaves the client’s network. In this paper, we present our design and show how it can be transparently integrated with the current Web. The paper is organized as follows. Section 2 gives a brief overview of the existing redirection methods for the World-Wide Web. It outlines the advantages and disadvantages of each method. Section 3 presents the principle of our distributed redirection scheme. Section 4 details the de- sign of the redirection server. Section 5 describes aspects concerning client and redirection server interaction. Section 6 presents simulation results of the distributed redirection mechanisms. Sec- tion 7 discusses the benefits and drawbacks of our redirection scheme with respect to other Web redirection mechanisms. And finally, Section 8 concludes and gives some future work directions. 1