The multicast bandwidth advantage in serving a web site

Abstract. Delivering popular web pages to the clients results in high bandwidth and high load on the web servers. A method to overcome this problem is to send these pages, requested by many users, via multicast. In this paper, we provide an analytic criterion to determine which pages to multicast, and analyze the overall saving factor as compared with a unicast delivery. The analysis is based on the well known observation that page popularity follows a Zipf-like distribution. Interestingly, we can obtain closed-form analytical expressions for the saving factor, that show the multicast advantage as a function of the site hit-rate, the allowed latency and the Zipf parameter. 1 Introduction One of the largest problems in the web is to deliver the content efficiently from the site to the user. High load on the server and on the network leads to long delays or more extremely denial of services. Increasing the capacity for delivering the content results in a high cost of extra servers and extra bandwidth. Moreover, the capacity is planed to some value, though larger than the average load, but almost always cannot accommodate the peak load. This is specially correct for popular pages were the access pattern may be unpredictable and very unstable (e.g. the famous Starr report case). There are several methods to try to overcome the problem. One is to use caches [13, 7, 1]. However, caches are not effective for frequently changing content or for long files (e.g video, audio). A different possibility that we consider in this paper is to use multicast [4, 8, 6], i.e., to deliver the content simultaneously to many (all) users via multicast dynamic tree. Obviously, one may also combine both caching and multicasting to further improve the solution