1.1 Current Development
In client-server network architecture, each computer acts as either a client or server. Client-server architecture illustrates the relationship between two computers in which one is a client, which makes service requests to another computer, a server. The upside of client-server architecture is easy to handle and extend its scale. However, with the increase of the requests by the clients, the server may be soon overloaded, and unable to deliver or poorly deliver the requested data.
In contrast to client-server architecture, peer to peer (P2P) architecture illustrates another relationship between two computers in network architecture. In recent years, P2P networks and P2P applications have been emerging as one of the most promising approaches to address the scalability problem. Unlike client-server networks, where network information is stored on a centralized server and made available to tens, hundreds, or thousands of clients, in P2P network, all peers have equivalent function, they share a part of their own resource such as storage capacity, file content, computing power to other peers. In other words, in P2P network, all peer acts as both a client and a server at the same time and collaborate with each other instead of relying on centralized servers.
It is obvious that P2P network is good in scalability. Beside it provides for peers an easy way to share and exchange information without the need of costly and difficult to maintain the central server. The first P2P application is Naptster [1], that was used for mp3 music file sharing. Napster totally overthrows the traditional client-server network architecture. After that, P2P-related applications for file sharing develop rapidly. ezPeer [2], eDonkey [3], eMule [4], Kazaa [5] and BitTorrent [6] are very popular file sharing software at present.
In addition to file sharing, there are some P2P applications also applied in other domain such as telephony and video streaming. Skype [7], a famous P2P telephony application, routes voice packets from callers to callees over P2P network, and allows
users to make phone calls over the Internet. Skype realizes real-time voice communication over P2P network.
1.2 Motivation
P2P streaming is another emerging domain on P2P applications, such as PPStream [8], PPlive [9], etc. In general, P2P streaming system can be categorized into two basic typologies: (1) In P2P live streaming systems, users choose from different channels and watch the live content together. In other words, users claim synchronous video content, in order to maintain the playback quality. Reducing end-to-end delay is an important issue in P2P live streaming system. (2) In P2P VoD (Video-on-Demand) streaming systems, the users may choose a specific movie and play it at any time because the video has been recorded in advance.
However, these two streaming system do not provide users watch time-shift streaming (or called live-shift streaming). In time-shift streaming systems, users can watch a live streaming with arbitrary offset of time, i.e., users can watch a live program from the beginning even when this live program has already started playing for a while. For example, someone likes to watch a ball game from the beginning but he misses the start time of the game, or someone likes to replay some spectacular moments in a program. These are very practical situations but there is no P2P streaming system that can provide time-shift streaming at present. That is because there are some difficult problems in developing and designing a P2P time-shift streaming system.
Since in both P2P VoD and time-shift streaming, users play a program at any time, the video content they claim are asynchronous, we can treat P2P time-shift streaming as a special case of P2P VoD system. The main difference between these two streaming systems is the length of video. In P2P VoD streaming system, the length of video are always fixed, but in P2P time-shift streaming, the length of video can be very long. Obviously, if a time-shift streaming system wants to provide users watch arbitrary offset of live content, storage policy for time-shift video content is a serious problem.
Although there are many studies about P2P live streaming and P2P VoD streaming, quite few studies can support P2P live and time-shift streaming at the same time. LiveShift [10], P2TSS [11] and IPTV [12] are the only three studies mention about time-shift streaming, but all of the three studies are just prototypes and have not been implemented. On the other hand, studies about P2P VoD are very plentiful.
Huang analyzes Microsoft client-server VoD system and outcomes that P2P VoD system can decrease the overhead of the VoD server. Because in P2P VoD streaming system, storing vast video content by using the storage of peers, and enabling peers to share their stored video content instead of always requesting video content from VoD servers. To conclude our above discussion, it is worth to design a P2P time-shift streaming system, using P2P network architecture to decrease the overhead of a dedicated time-shift server and to provide time-shift streaming and live streaming at the same time.
1.3 Objective
In this thesis, we propose a P2P time-shift streaming system which can play live streaming and time-shift streaming. Although P2P time-shift streaming is different from P2P live and P2P VoD streaming functionally, by studying documents of P2P live streaming and P2P VoD streaming, we conclude that our P2P time-shift streaming system should cope well with the following issues:
(1) Storage and search policy of the time-shift streaming content.
(2) Management of live and time-shift group.
(3) Transmission architecture for live streaming and time-shift streaming.
(4) Prefetch policy for time-shift streaming content.
These four issues are associated with storage policy. Therefore, the storage policy is the most important in our system. We propose an effective storage policy that not only decides how to use storage efficiently and what video content should be stored but also helps peers sharing and delivering their video content efficiently.
Our P2P time-shift streaming system has been designed based on our proposed storage policy. We believe our work is valuable since there are no existing P2P streaming systems that can actually provide time-shift and live streaming at the same time.
1.4 Summary
The remaining part is organized as follows. Chapter 2 describes the current work in P2P streaming researches related to our system. Chapter 3 shows our system design in details. Chapter 4 presents the system implementation and system analyzsis. Finally, we give our conclusions in Chapter 5.