A Scaleable Multicast Protocol to Enable Future-Generation Large Scale Multimedia Applications

点击翻译按钮获取中文摘要

A picture is worth a thousand words, and the addition of sound and motion can breathe life into a picture.As a result, the use of video in all kinds of multimedia applications has become an important developmentin the 1990s. In particular, Video on demand (VOD) is a core technology for many important applicationssuch as digital libraries, distance learning, electronic commerce, entertainment, public informationsystems, etc.The simplest video delivery technique employs a dedicated stream for each service request. Obviously,this scheme is too expensive and has been shown to have little scalability. To vastly reduce this cost, onecan leverage multicast technology to allow multiple clients to share a video stream. Unfortunately, today'smulticast technology was developed in the 80's, and was not optimized for such applications. Missing amulticast could mean a long wait until the next multicast. This limitation has recently led to a large bodyof research looking for remedies at the application level.Application-level solutions, however, cannot address all the drawbacks, particularly traffic congestioncaused by too many multicasts. In fact, this proposal is motivated by application-level work done bythe investigator in his last NSF project (from 1998 to 2000). The achievements of that project, thoughsignificant, were constrained by the limitation of standard multicast. Minimizing the multicast frequencyis crucial for any multicast-based applications, particularly those running on a public network such as theInternet. Attaining this goal requires a fundamental change in the multicast concept. The investigatorproposes enabling the multicast tree to deliver the entire video to all the receivers who may subscribe tothe multicast at different times. This new capability would significantly reduce the number of multicastsnecessary, and therefore lower the network resource requirements. At first sight, this seems like a missionimpossible. But on the contrary, this can be achieved by allowing the routers on the multicast tree tointercept and cache a video stream passing through. The data can be relayed to subsequent subscribers tofurther extend the multicast tree. In this project, the investigator proposes to develop a suite of techniquesto make such an environment practical for future-generation large-scale multimedia applications.To assess the potential of the proposed idea, the investigator has designed an initial version of hisCaching Multicast Protocol (CMP). The preliminary study indicates that this approach has the potentialto revolutionize the way in which multimedia information is disseminated. This scheme is much morescalable and less expensive than application-level solutions. The new communication paradigm, however,demands new research at the network level. In this project, the investigator proposes to study issues suchas cache organization, cache replacement policies, multicast routing algorithms, congestion control, clientand server protocols, etc. In addition, he will implement a prototype, consisting of CMP-capable routers,CMP-compliant servers (sources) and CMP-compliant clients (destinations) to examine the practicality ofthe proposed techniques.The significance of this project is in advancing multicast technology to enable large-scale deploymentof multimedia applications. Although many great advances have been made in the field of informationtechnology, none have had as great and direct an impact on the daily lives of ordinary citizens as multime-diainformation systems. As such, the benefits of this research will reach many people in society. In termsof education, this project will help to support three doctoral students. The investigator has a good recordof using NSF projects to train students for teaching careers. After these students graduate, they can helpto spread the CMP expertise developed from this project to their new institutions. Other graduate studentstaking our advanced databases course can also benefit from this work by experimenting with various newmulticast applications on the CMP testbed. To reach out to members of communities, materials resultingfrom this research will be disseminated on the investigator's research group website.