The recent success of multicast applications such as Internet teleconferencing illustrates the tremendous potential of applications built upon wide-area multicast communication services. A critical issue for such multicast applications and the higher layer protocols required to support them is the manner in which packet losses occur within the multicast network. In this paper we present and analyze packet loss data collected on multicast-capable hosts at 17 geographically distinct locations in Europe and the US and connected via the MBone. We experimentally and quantitatively examine the spatial and temporal correlation in packet loss among participants in a multicast session. Our results show that there is some spatial correlation in loss among the multicast sites. However, the shared loss in the backbone of the MBone is, for the most part, low. We find a fairly significant amount of of burst loss (consecutive losses) at most sites. In every dataset, at least one receiver experienced ...

This article describes the network concepts underlying MBone, the importance of bandwidth considerations, various application tools, MBone events, interesting MBone uses (see the two sidebars), and provides guidance on how to connect your Internet site to the MBone.

In this paper we consider the problem of choosing among a collection of replicated servers, focusing on the question of how to make choices that segregate client/server traffic according to network topology.We explore the cost and effectiveness of a variety of approaches, ranging from those requiring routing layer support (e.g., anycast) to those that build location databases using application-level probe tools like traceroute. Weuncover a number of tradeoffs between effectiveness, network cost, ease of deployment, and portability across differenttypes of networks. We performed our experiments using a simulation parameterized by a topology collected from 7 survey sites across the United States, exploring a global collection of Network Time Protocol servers.

The IP multicast delivery mechanism provides a popular basis for delivery of continuous media to many participants in a conferencing application. However, the best-effort nature of multicast delivery results in poor playback quality in the presence of network congestion and packet loss. Contrary to widespread belief that the real-time nature of continuous media applications precludes the possibility of recovery of lost packets using retransmissions, we have found that these applications offer an interesting tradeoff between the desired playback quality and the desired degree of interactivity. In particular, we propose a new model of multicast delivery called resilient multicast in which each receiver in a multicast group can decide its own tradeoff between reliability and real-time requirements. To be effective, error recovery mechanisms in such a model need to be both fast (due to the real-time constraint) and have a low overhead (due to high volume of continuous media data). We have...

Network support for multicast has triggered the development of group communication applications such as multipoint data dissemination and multi-party conferencing tools. To support these applications, several multicast transport protocols have been proposed and implemented. Multicast transport protocols have been an area of active research for the past couple of years. This document tries to summarize the activities in this work-in-progress area by surveying several multicast transport protocols. The paper also presents a taxonomy to classify the surveyed protocols according to several distinct features, discusses the rationale behind the protocol's design decisions, and presents some current research issues in multicast protocol design. 1 Introduction Multicast transport mechanisms have been a topic of intense research and development efforts over the past couple of years. Both the Internet Engineering and Internet Research Task Forces (IETF and IRTF) have been heavily involved in co...

The recent growth in use of the World-Wide Web in the Internet has caused a significant increase in the demand placed on Web servers. This increased load results in noticeably longer response times for users. We propose an approach to using multicast in the delivery of Web resources that reduces the load on servers as well as the networks that connect them. We analyze the issues involved in using multicast in the Web, especially those related to routing and addressing. We also discuss the design and implementation of a system based on the existing WWW client and server architecture and the multicast support provided within IP. College of Computing Georgia Institute of Technology Atlanta, GA 30332-0280 rjc@cc.gatech.edu This research is supported by a grant from the National Science Foundation (NCR-9305115) and ARPA contract N00174-93-K-0105. 1 Introduction The number of World-Wide Web users in the Internet has grown dramatically in recent months. During the period from September t...

The aim of the MICE project is to pilot multimedia interworking between European researchers, using a heterogeneous hardware platform and existing network facilities. Part of the project brief was to assess cost and benefit of providing a regular multimedia conferencing service for research collaboration, and provide recommendations for the deployment and use of such systems in future. In order to gain gain hands-on experience of both running and using such a service, MICE partners started a distributed International Research Seminar Series, in which researchers and students could participate. Speakers and audiences participated from conference rooms and workstations at MICE partner sites, and increasingly at other remote sites in Europe, the US and Australia. Observations, recordings and comments provided data on which the cost-benefit analysis and recommendations for such a service are based.

Current and future Internet services will provide a large, rapidly evolving, highly accessed, yet autonomously managed information space. Internet news, perhaps, is the closest existing precursor to such services. It permits autonomous updates, is replicated at thousands of autonomously managed sites, and manages a large database. It gets its performance through massive replication. This paper proposes a scalable mechanism for replicating wide-area, autonomously managed services. We target replication degrees of tens of thousands of weakly consistent replicas. For efficiency, our mechanism probes the network and computes a good logical topology over which to send updates. For scalability, we organize replicas into hierarchical replication groups, analogous to the Internet's autonomous routing domains. We argue that efficient, massive replication does not have to rely on internet multicast.

Straightforward, one-way delivery of audio/video through television sets has existed for many decades. In the 1980s, new services like pay-per-view and video-on-demand were touted as the "killer applications" for interactive TV. However, the hype quickly died away, leaving only hard technical problems and costly systems. As an alternative, we propose a new jukebox paradigm offering flexibility in how programs are requested and scheduled for playout. The jukebox scheduling paradigm offers flexibility ranging from complete viewer control (true video-on-demand), to complete service provider control (traditional broadcast TV). In this paper, we first describe our proposed jukebox paradigm and relate it to other on-demand paradigms. We also describe several critical research issues, including the one-to-many delivery of content, program scheduling policies, server location, and the provision of advanced services like VCR-style interactivity and advanced reservations. In addition, we present our implementation of a jukebox-based service called the Interactive Multimedia Jukebox(IMJ). The IMJ provides scheduling via the World Wide Web (WWW) and content delivery via the Multicast Backbone (MBone). For the IMJ, we present usage statistics collected during the past couple of years. Furthermore, using this data and a simulation environment, we show that jukebox systems have the potential to scale to very large numbers of viewers.

nt and migration. At the application layer, the existence of network multicast does not guarantee that an application can take advantage of it. We demonstrate the adaptation of a traditionally unicast application -- database sharing -- to multicast, thus achieving scaling in network and processor performance. For the transport layer, there is a growing consensus in the research community that relying on generic protocols can significantly reduce the efficiency of many applications. This effect seems to be even more pronounced for multicast transport protocols; consequently, protocols may need to be constructed for different types of applications iii to achieve increased efficiency. We propose a reliable multicast transport protocol tuned specifically for bulk-data applications, such as our multicast-adapted database sharing application. In developing this new transport protocol, we employ a synergy of new and existing techniques in multicast transport. We propose a new sender s