P2PVR: A playback offset aware multicast tree for on-demand video streaming with VCR functions

https://doi.org/10.1016/j.sysarc.2011.03.001Get rights and content

Abstract

Tree-based peer-to-peer multicast overlays have been widely proposed to provide on-demand video streaming services. However, alleviating the server load in large-scale, dynamic environments such as the Internet remains a major challenge. Accordingly, this paper proposes a new peer-to-peer on-demand video streaming system, designated as peer-to-peer personal video recorder (P2PVR), which facilitates essential VCR functions in an efficient manner. In the proposed architecture, the peers are organized into a playback offset aware tree-based overlay, in which they share streaming data with other peers having a similar playback offset. In addition, a semi-decentralized directory service is developed to assist peers in searching only those parts of the tree known to contain eligible parent nodes, i.e. nodes which possess the expected streaming data. The performance of P2PVR is evaluated by performing a series of experiments on the PlanetLab platform. The results show that P2PVR yields a significant reduction in the server stream stress and achieves a reasonably low playback discontinuity under dynamic network conditions whilst simultaneously granting a low control overhead and startup latency.

Introduction

Peer-to-peer (P2P) networking has emerged as a promising solution for the provision of video streaming services due to its flexibility and readily-deployable nature. Existing P2P approaches can be classified as either tree-based [1], [2], [3], [4], [5], [6], [7], [22], [24] or mesh-based [8], [9], [10], [11], depending on the distribution graph they implement. In mesh-based overlays, each peer advertises the packets within its possession to its neighbors, and the neighbors explicitly pull those packets which they require. However, the advertising and pull processes result in a large overhead or long startup delays, significant video switching delays and large peer playback time lags [13], [14], [15], [16]. For example, Huang et al. [25] reported that the average control overhead of PPLive [26], a commercial P2P video streaming system using mesh-based overlays, is around 10%. Tree-based overlays, by contrast, implement a tree distribution graph rooted at the source of the content and deliver data packets along this tree without any explicit requests. Thus, tree-based approaches do not incur the control overhead vs. delay tradeoff problem inherent in mesh networks and are therefore more suitable for applications with low bandwidth and minimal delay requirements. However, complex distributed algorithms are required to construct and maintain the tree overlay as new peers join the network or existing peers leave. Furthermore, if these algorithms are unable to locate eligible parent peers (i.e., peers with the expected streaming data), the requesting peer has no choice but to join the source server. In large-scale, dynamic environments, the server is soon overwhelmed by the sheer volume of streaming requests, and thus all the peers experience discontinuous playback. As a result, reducing the server stream stress is a critical issue in the design and implementation of tree-based P2P networks.

In on-demand video streaming systems, locating eligible parent peers is highly challenging due to the constant changes which take place in the network bandwidth and user dynamics. As a result, some form of directory service is required to provide indexing information relating the network addresses of all the peers in the system to the data cached within their buffers. In Ref. [1], the directory service was implemented using a single server, and was therefore vulnerable to the single point of failure problem. In Refs. [5], [6], the control overhead incurred by the directory service was distributed over all the peers in the system via the use of complicated data structures. However, constructing and maintaining these structures imposed a heavy load on the network. In some studies [2], [3], [4], [7], [24], the directory service is distributed over both the server and all the peers in order to balance the control overhead imposed on the server and the network, respectively. In such systems, the existing peers are divided into multiple groups, and the requesting peers search only one or two of these groups to locate eligible parent nodes. However, the overlay construction and maintenance mechanisms used in these systems do not make full use of the directory service, and thus some groups within the network containing eligible peers are overlooked. As a result, the upload capacities of these ‘hidden’ peers are underutilized; obliging many peers to connect directly to the server. Therefore, a requirement exists for scalable streaming systems capable of providing an acceptable streaming quality under even highly dynamic network environments whilst simultaneously maintaining a low control overhead and startup latency.

Accordingly, this study proposes a scalable and efficient P2P architecture designated as peer-to-peer personal video recorder (P2PVR) for the provision of on-demand video streaming with VCR functions over the Internet. In the proposed architecture, the peers are organized into a playback offset aware multicast tree, rooted at the server. In constructing the tree, the peers with larger playback offsets are located closer to the root node since, by definition, they possess the contents of interest to peers with smaller playback offsets. The streaming data is cached in accordance with the buffer capacities of the peers and is relayed from the higher-level nodes to the lower-level nodes without the need for explicit pull messages. In addition, a semi-decentralized directory service is proposed to help requesting peers locate suitable parent nodes. Importantly, the directory service enables the peers to search only those parts of the tree known to contain eligible parent nodes. In other words, the proposed service not only eases the server load, but also reduces the search time and control overhead.

The objective of the P2PVR system proposed in this study is to provide a practical streaming service capable of accommodating the highly dynamic nature of the Internet. Thus, the performance of the proposed system is evaluated by performing a series of experiments on PlanetLab [17]; a testbed overlaid on the Internet. The results show that P2PVR has good scalability and accommodates highly dynamic client behaviors without the need for a large buffer space. In addition, it is shown that the system achieves a reasonably low playback discontinuity under different client configurations whilst maintaining a low control overhead and startup latency.

The remainder of this paper is organized as follows. Section 2 reviews the related work. Section 3 describes the architecture of the proposed P2PVR system and introduces the related overlay construction and maintenance mechanisms. Section 4 describes the experimental methodology used to evaluate the performance of P2PVR. Section 5 presents and discusses the experimental results. Finally, Section 6 summarizes the major contributions of the study and indicates the intended direction of future research.

Section snippets

Related work

Many methods have been proposed over the past decade for the provision of on-demand video streaming services using tree-based P2P overlays [1], [2], [3], [4], [5], [6], [7], [24]. Among these systems, CoopNet [1] is intended primarily for live video streaming, but also supports on-demand video streaming. In CoopNet, the media content is encoded into several streams using a multiple description coding (MDC) technique [18], and the streams are then transmitted over several different distribution

Peer-to-peer personal video recorder

This section describes the P2PVR architecture proposed in this study for providing a scalable and efficient solution for the provision of on-demand video streaming services with VCR functions. The major notations used throughout the remainder of this paper are summarized in Table 1.

PlanetLab environment

The performance of P2PVR was evaluated by conducting a series of experiments on the PlanetLab platform. PlanetLab is a global research network consisting of approximately 1132 nodes located around the world. To understand the characteristics of the network paths connecting the PlanetLab nodes, preliminary pair-wise latency and available bandwidth measurement tests were conducted on 416 nodes on February 18th, 2009. In the tests, each node probed the other nodes sequentially while being

Performance evaluation in stable environment

In the first set of experiments, all of the nodes remained in the network once they had joined, and played the video continuously from the beginning to the end. Each experiment was run for a total of 30 min.

Conclusion

This paper has presented a playback offset aware tree-based overlay system designated as P2PVR for performing on-demand video streaming with VCR functions. The key design issues of P2PVR have been discussed, and a semi-decentralized directory service for overlay construction and maintenance has been proposed. The performance of P2PVR has been evaluated under real-world network conditions using the PlanetLab platform. The experimental results have shown that P2PVR delivers an acceptable

Yun-Shuai Yu obtained his B.S. and M.S. degrees in electrical engineering from National Cheng Kung University, Taiwan, in 2002 and 2004, respectively. He is currently working towards the Ph.D. degree in the Department of Electrical Engineering at National Cheng-Kung University, Taiwan. His research interests include P2P computing, streaming, computer networking, and embedded system.

References (31)

  • T.T. Do et al.

    Robust video-on-demand streaming in peer-to-peer environments

    Computer Communications

    (2008)
  • Y. Guo et al.

    DirectStream: a directory-based peer-to-peer video streaming service

    Computer Communications

    (2008)
  • B. Cheng et al.

    Evaluation and optimization of a peer-to-peer video-on-demand system

    Journal of Systems Architecture

    (2008)
  • V.N. Padmanabhan, H.J. Wang, P.A. Chou, K. Sripanidkulchai, Distributing streaming media content using cooperative...
  • Y. Guo, K. Suh, J. Kurose, D. Towsley, P2Cast: peer-to-peer patching scheme for VoD service, in: Proceedings of WWW’03,...
  • T.T. Do, K.A. Hua, M.A. Tantaoui, P2VoD: providing fault tolerant video-on-demand streaming in peer-to-peer...
  • R. Zhang, A.R. Butt, Y.C. Hu, Topology-aware peer-to-peer on-demand streaming, in: Proceedings of 2005 IFIP Networking...
  • Y. He et al.

    VOVO: VCR-oriented video-on-demand in large-scale peer-to-peer networks

    IEEE Transactions on Parallel and Distributed Systems

    (2009)
  • M. Zhou et al.

    Tree-assisted gossiping for overlay video distribution

    Multimedia Tools and Applications

    (2006)
  • B. Cheng, H. Jin, X. Liao, Supporting VCR functions in P2P VoD services using ring-assisted overlays, in: Proceedings...
  • W.-P.K. Yiu et al.

    VMesh: distributed segment storage for peer-to-peer interactive video streaming

    IEEE JSAC

    (2007)
  • D. Wang et al.

    A dynamic skip list-based overlay for on-demand media streaming with VCR interactions

    IEEE Transactions on Parallel and Distributed Systems

    (2008)
  • A. Sentinelli et al.

    Will IPTV ride the peer-to-peer stream?

    IEEE Communications Magazine

    (2007)
  • M. Zhang et al.

    Understanding the power of pull-based streaming protocol: can we do better?

    IEEE JSAC

    (2007)
  • V. Venkatramen, K. Yoshida, P. Francis, Chunkyspread: heterogeneous unstructured tree-based peer-to-peer multicast, in:...
  • Cited by (11)

    View all citing articles on Scopus

    Yun-Shuai Yu obtained his B.S. and M.S. degrees in electrical engineering from National Cheng Kung University, Taiwan, in 2002 and 2004, respectively. He is currently working towards the Ph.D. degree in the Department of Electrical Engineering at National Cheng-Kung University, Taiwan. His research interests include P2P computing, streaming, computer networking, and embedded system.

    Ce-Kuen Shieh obtained his B.S., M.S., and Ph.D. degrees in electrical engineering from National Cheng Kung University, Taiwan, in 1977, 1983, and 1988, respectively. Currently, he is a professor in the Department of Electrical Engineering at National Cheng Kung University, Taiwan. His research interests include distributed and parallel processing, operating systems, computer networking, and compiler.

    Chun-Hsiang Lin obtained his B.S. and M.S. degrees in electrical engineering from National Cheng Kung University, Taiwan, in 2005 and 2007, respectively. He is currently a member of technical staff at ASUS Inc., Taiwan. His research interests include P2P computing and streaming.

    Szu-Yu Wang obtained his B.S. and M.S. degrees in electrical engineering from National Cheng Kung University, Taiwan, in 2007 and 2009, respectively. He is an engineer of Android engineering at HTC Corporation, Taiwan. Currently he is developing mobile application for Open Mobile Alliance Device Management. His research interests include P2P computing and streaming.

    View full text