Conference Paper

Storage optimization for a peer-to-peer video-on-demand network

DOI: 10.1145/1730836.1730844 Conference: Proceedings of the First Annual ACM SIGMM Conference on Multimedia Systems, MMSys 2010, Phoenix, Arizona, USA, February 22-23, 2010
Source: DBLP


This paper explores requirements for efficient pre-seeding of video-on-demand (VoD) movie data onto numerous customer set-top boxes in a cable ISP environment. The pre-seeded content will then be distributed to other set-top boxes in the same cable community using a peer-to-peer (P2P) network protocol such as BitTorrent. The challenges and solutions required for P2P VoD provided by a fixed provider such as a cable company are fundamentally different from those seen in traditional P2P networks or client-server VoD solutions. Our work pre-positions data into set-top boxes using a mathematical programming algorithm. The objective of the algorithm is to minimize uplink traffic, given a popularity model for various pieces of content and information about storage and bandwidth capacity constraints at the customer nodes. Given the complex non-linear nature of P2P interactions, these mathematical programs are solved using non-linear optimization approaches. Using a BitTorrent-like peer-to-peer data delivery system, we show through extensive simulations that our mathematical model for pre-seeding data based on object popularity and node bandwidth availability leads to noticeably greater reductions in uplink traffic and VoD server load than a weighted-random pre-seeding scheme that only considers object popularity.

13 Reads
  • Source
    • "The peers help to reduce the traffic in the network core, but the servers still remain an inevitable part of the system that guarantees the required QoS. These solutions together with various content distribution schemes prove to be efficient in reducing [4], [5], [6], [7], [8], [9] and even eliminating the traffic requested from the streaming servers [10], [11]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Video on Demand is a leading TV service offered by the IPTV operators in the past decade that has been rapidly gaining on popularity because it offers great convenience to the customers to watch any video they want at any time. However, the drawback of this service it that it is very resource demanding and expensive for the operator. One of the solutions for reducing the high traffic demands from the video servers is the imple-mentation of peer-assisted streaming, i.e., including the peers in the streaming process by taking advantage of their unused streaming and storage capacity. In order to estimate the reduction of the traffic demand depending on the network configuration and the intensity of demands for videos, we developed a stochastic model that determines the system behavior in stationary state. We proved that the model is a precise tool for estimating the server demands. However, the model requires high computation power for obtaining the desired results. The size of the number of linear equations that have to be solved grows rapidly with the growth of the number of peers and their streaming capacity, which is serious issue for using the model as a tool for estimation of the system performance. Therefore, we propose a sampling method that significantly reduces the size of the system of linear equations, and thus, reduces the computation time and resources required for obtaining the results. Our analysis shows that although the size of the system is reduced, the relative error compared to the original unreduced system is negligible.
    CIIT; 04/2014
  • [Show abstract] [Hide abstract]
    ABSTRACT: Recently, a new peer-assisted architecture to build content delivery systems has been presented. This architecture is based on the use of the storage capacity of end-users’ set-top boxes (STBs), connected in a peer-to-peer (P2P) manner in order to help the content servers in the delivery process. In these systems, the contents are usually split into a set of smaller pieces, called sub-streams, which are randomly injected at the STBs. The present paper is focused on Video on Demand (VoD) streaming and it is assumed that the STB-based content delivery system is deployed over the global Internet, where the clients are distributed over different ISP networks. In this scenario, three different strategies are studied for increasing the percentage of data uploaded by peers, in order to offload the content servers as much as possible. First of all, a new mechanism is presented which determines which sub-stream has to be placed at which STB by a Non-Linear Programming (NLP) formulation. A different strategy for reducing the content server load is to take advantage of the available bandwidth in the different ISP networks. In this sense, two new mechanisms for forwarding the VoD requests to different ISP networks are presented. Finally, the present paper also shows that in some situations the available uplink bandwidth is associated with STBs that do not have the required sub-streams. Regarding this concern, a new mechanism has been designed that dynamically re-allocates some streams, which are being transmitted from specific STBs, to different STBs, in order to find the necessary resources to start new streaming sessions.
    Peer-to-Peer Networking and Applications 09/2012; 6(3). DOI:10.1007/s12083-012-0174-2 · 0.63 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: With the rise of VOD (Video-on-Demand) services provisioning as a successful service on the Internet and managed networks, we are witnessing a drive towards cost-efficiency and economies of scale. Many broadband operators around the world are experimenting with P2P (Peer-to-Peer) systems centered on STBs (Set-Top-Boxes) to increase the competitiveness of their VOD services offering. By leveraging the storage and uplink bandwidth capacities available at a certain number of STBs operated by the broadband operator, the savings in terms of backend streaming capacities will represent sizable and decisive gains in cost. In these systems, video contents are usually fragmented into a number of complementary content fragments, called sub-streams, which are randomly injected in the network of STBs, and the VOD service is essentially provisioned through multisource streaming sessions from neighboring STBs to the requesting STB. One of the main challenges in such peer-assisted streaming systems remains the maximization of the utilization of STB resources utility for a given content popularity pattern. In this paper, we specifically focus on the content injection strategy and how the different content fragments should be dispatched in the network to achieve the highest performance in the VOD services provisioning epoch. We demonstrate that the random injection strategy is not appropriate for maximizing the number of simultaneous VOD streaming sessions in the network. Our objective is to first gain a better understanding of the factors driving P2P-based VOD streaming systems and provide guidelines to better operate such systems and ultimately give service operators the tools to achieve different performance objectives and/or fit specific network configurations. Further, we propose a new content dispatching strategy that maximizes the number of served VOD sessions by balancing the streaming load among the different STBs. Finally, we propose a complementary streaming resources reprovisioning mechanism that acts in real-time to reprovision the resources for serving VOD sessions to new STBs and to release trapped resources for new incoming VOD service requests.
    Multimedia Tools and Applications 06/2012; 70(3). DOI:10.1007/s11042-012-1171-4 · 1.35 Impact Factor
Show more


13 Reads
Available from