Reliable Multicast Protocol Specialization for Caching and Collaboration within the World-Wide Web

The World Wide Web (WWW) has become an important medium for information dissemination. One model for synchronous information dissemination is a scheme called webcasting where data is simultaneously distributed to multiple destinations. The WWW's traditional unicast client/server communication model su ers, however, when applied to webcasting; solutions which require many clients to simultaneously fetch data from the origin server using the client/server model will likely cause server and link overload. A number of webcasting solutions have been proposed. Many have limited scalability because they are based on unicast while others use multicast for more scalable data delivery but require server modi cation or have rigid architectures. We believe that successful webcasting solutions will provide scalable, reliable delivery yet still be compatible with the existing Web infrastructure. In this paper we describe a webcast design which improves upon previous designs by leveraging the application level framing (ALF) design methodology. We build upon the Scalable Reliable Multicast (SRM) framework, which is based upon ALF, to create a custom protocol to meet webcasting's needs: reliable delivery which is scalable with respect to both origin server load and link load. We employ the protocol in a webcasting architecture consisting of two reusable components, a webcache component and a browser control component. The resulting application is a scalable webcasting application that requires no browser modi cations or proxy speci cation. We have implemented our design in a webcast application called MASHCast. Initial measurements showed that our scalable custom protocol reduced load times from 7% to 53% depending upon the size of the page and the round-trip time to the page's origin. This work is supported by the Advanced Research Projects Agency (N66001-96-C-8508) and California MICRO.