Video transmission over IP is currently a hot topic both in entertainment and research communities. A problem that threatens the development of video over IP services is the bandwidth required to serve a potentially very large number of users. In this context, Peer-to-peer (P2P) technologies are considered a possible solution for the distribution of video content to many users. This chapter describes a novel P2P transport protocol suited for live multimedia streaming. The described protocol has low start-up time, it is robust with respect to data losses (due to congestion or node departure) and it can help counteracting the malicious injection of “bogus packets” in the media stream. The proposed protocol can be used with any type of data and, from the application point of view, it appears as a protocol similar to TCP or UDP, making the reuse of existing software and protocols easier.
Top1. Introduction
Because of its potentially large user base, Internet is currently considered to be an interesting platform for implementation and deployment of innovative video services. Although a large audience is clearly what a video service provider desires, serving a large number of users requires powerful bandwidth resources1 to be allocated, on a worst-case basis, to the server(s). Moreover, the bandwidth problem, already a serious issue with video media, will get even worse with new emerging type of multimedia such as 3D data. A possible solution to the bandwidth problem could be the use of IP multicast. Unfortunately, IP multicast is not a universally feasible solution since it requires multicast-enabled routers which are not widely deployed.
Peer-to-peer (P2P) systems are considered a promising solution for video streaming to a large number of users (Alstrup & Rauhe, 2005, Kozamernik, 2000, Fodor & Dán, 2007, Padmanabhan, Wang, Chou, & Sripanidkulchai, 2002). The reason for this interest is that each new node contributes with its own resources by forwarding the received data to other nodes. This means that although the download bandwidth increases linearly with the number of users, also the available upload bandwidth increases with the number of users. Ideally, if each user were able to provide an upload bandwidth equal to the download one, the server would need only to “seed” few nodes of the network, and the network would take care of the distribution by itself.
What makes media streaming over P2P networks an active field of research is that multimedia streaming (and, in particular, live event streaming) has requirements that are quite different from the requirements of the more common P2P file-sharing applications (e.g., Napster, Gnutella, KaZaA, BitTorrent). Maybe the most important difference is that while in a file-sharing system there is no time constraint, in a streaming context, packets have an expiration date and if a node misses a few packets, the user can experience an unacceptable drop in quality. This problem, together with other problems related to P2P streaming (such as the asymmetric bandwidth problem or the possibility of stream poisoning, analyzed in detail in Section 2.3) makes multimedia streaming a peculiar applicative context for P2P solutions.
This chapter has the following structure. First, in Section 2 we review the main P2P structures, both for file sharing and multimedia streaming, and the requirements that a P2P system for media streaming must satisfy. Successively, in Section 3 we describe a new P2P transport protocol, suited for live multimedia streaming, based on the chunkless approach described in (Bernardini, Rinaldo, & Vitali, 2008) and currently developed as part of the SourceForge project corallo2. The described protocol does not require buffering and is robust with respect to data losses (due to congestion or node departure) and stream poisoning attacks. Although the protocol was developed for streaming applications, it makes no assumption on the format of transported data, so it can be used with any type of data (audio, video, documents, …).