[Show abstract][Hide abstract] ABSTRACT: BitTorrent is one of the Internet’'s most efficient content distribution protocols. It is known to perform very well over the wired Internet where end-to-end performance is almost guaranteed. However, in wireless ad hoc networks, many constraints appear as the scarcity of resources and their shared nature, which make running BitTorrent in such an environment with its default configuration not lead to best performances. To these constraints it adds the fact that peers are both routers and end-users and that TCP-performance drops seriously with the number of hops. We show in this work that the neighbor selection mechanism in BitTorrent plays an important role in determining the performance of the protocol when deployed over a wireless ad hoc network. It is no longer efficient to choose and treat with peers independently of their location. A first solution is to limit the scope of the neighborhood. In this case, TCP connections are fast but there is no more diversity of pieces in the network: pieces propagate in a unique direction from the seed to distant peers. This prohibits peers from reciprocating data and leads to low sharing ratios and suboptimal utilization of network resources. To recover from these impairments, we propose BitHoc, an enhancement to BitTorrent, which aims to minimize the time to download the content and at the same time to enforce cooperation and fairness among peers. BitHoc considers a restricted neighborhood to reduce routing overhead and to improve throughput, while establishing few connections to remote peers to improve diversity of pieces. To support this, BitHoc modifies the choking algorithm and adds a new piece selection strategy. With the help of extensive NS-2 simulations, we show that these enhancements to BitTorrent significantly improve the file completion time while fully profiting from the incentives implemented in BitTorrent to enforce fair sharing.
[Show abstract][Hide abstract] ABSTRACT: In this paper, we conduct extensive simulations to understand the properties of the overlay generated by BitTorrent. We start by analyzing how the overlay properties impact the efficiency of BitTorrent. We focus on the average peer set size (i.e., average number of neighbors), the time for a peer to reach its maximum peer set size, and the diameter of the overlay. In particular, we show that the later a peer arrives in a torrent, the longer it takes to reach its maximum peer set size. Then, we evaluate the impact of the maximum peer set size, the maximum number of outgoing connections per peer, and the number of NATed peers on the overlay properties. We show that BitTorrent generates a robust overlay, but that this overlay is not a random graph. In particular, the connectivity of a peer to its neighbors depends on its arriving order in the torrent. We also show that a large number of NATed peers significantly compromise the robustness of the overlay to attacks. Finally, we evaluate the impact of peer exchange on the overlay properties, and we show that it generates a chain-like overlay with a large diameter, which will adversely impact the efficiency of large torrents.
[Show abstract][Hide abstract] ABSTRACT: Network coding is a new transmission paradigm that proved its strength in optimizing the usage of net- work resources. In this paper, we evaluate the gain from using network coding for file sharing applica- tions running on top of wireless mesh networks. With extensive simulations carried out on a simulator we developed specifically for this study, we confirm that network coding can improve the performance of the file sharing application, but not as in wired networks. The main reason is that nodes over wireless cannot listen to different neighbors simultaneously. Never- theless, one can get more from network coding if the information transmission is made more diverse inside the network. We support this argument by varying the loss rate over wireless links and adding more sources.
[Show abstract][Hide abstract] ABSTRACT: Content distribution using the P2P paradigm has become one of the most dominant services in the Internet today. Most of the research effort in this area focuses on developing new distribution architectures. However, little work has gone into identifying the principle design choices that draw the behavior of the system. In this paper, we identify these design choices and show how they in uence the performance of different P2P architectures. For example, we discuss how clients should organize and cooperate in the network. We believe that our findings can serve as guidelines in the design of efficient future architectures.
Full-text · Article · Oct 2005 · ACM SIGCOMM Computer Communication Review
[Show abstract][Hide abstract] ABSTRACT: We study a new approach to provide an efficient VoD service to a large client population in P2P networks. Previous work has suggested to construct a multicast tree to distribute the video to clients. However, a multicast tree usually requires the central server to perform complex algorithms to add new clients to the tree. In this paper, we show how to simplify the algorithm performed by the server and, at the same time, achieve an efficient use of the system resources. For this purpose, we present a new pull-based approach, called PBA. The basic idea of PBA is quite simple. When a new client wishes to receive a video, the new client contacts first the server. If there is enough left over bandwidth along the path to the new client, the server transmits the video to the new client. Otherwise, the server provides to the new client a list of candidate servants chosen at random. These servants are clients that have received or are currently receiving the video. The new client then searches for an available servant to download the video from.
We compare PBA to P
2cast, a multicast-tree based approach proposed previously. We investigate via intensive simulations the efficiency of both approaches in terms of percentage of rejected clients and of bandwidth consumption. PBA does not only simplify the algorithm performed at the server, but also consumes less bandwidth and allows more clients to access the service.
[Show abstract][Hide abstract] ABSTRACT: Popular content such as software updates is requested by a large number of users. Traditionally, to satisfy a large number of requests, lager server farms or mirroring are used, both of which are expensive. An inexpensive alternative are peer-to-peer based replication systems, where users who retrieve the file, act simultaneously as clients and servers. In this paper, we study BitTorrent, a new and already very popular peer-to-peer application that allows distribution of very large contents to a large set of hosts. Our analysis of BitTorrent is based on measurements collected on a five months long period that involved thousands of peers. We assess the performance of the algorithms used in BitTorrent through several metrics. Our conclusions indicate that BitTorrent is a realistic and inexpensive alternative to the classical server-based content distribution.
[Show abstract][Hide abstract] ABSTRACT: Popular content such as software updates is requested by a large number of users. Traditionally, to satisfy a large number of requests, lager server farms or mirroring are used, both of which are expensive. An inexpensive alternative are peer-to-peer based replication systems, where users who retrieve the file, act simultaneously as clients and servers. In this paper, we study BitTorrent, a new and already very popular peerto -peer application that allows distribution of very large contents to a large set of hosts. Our analysis of BitTorrent is based on measurements collected on a five months long period that involved thousands of peers.
[Show abstract][Hide abstract] ABSTRACT: Cost-effectiveness is of foremost importance for large scale VoD systems. We assume a VoD system where each video is split
into two parts, the prefix and the suffix. We consider two new architectures: One architecture where the clients are equipped
with set-top boxes that allow to store locally the prefix part of some/all popular videos and second architecture where the
suffix is transmitted via satellite.
For each architecture, we develop a cost model to compute the delivery cost of videos. We show that these architectures are
efficient and significantly reduce the system cost in many scenarios: (i) By more than 45% with set-top boxes at the client
side, (ii) By more than 80% for satellite transmission of the suffix.
[Show abstract][Hide abstract] ABSTRACT: In this chapter, we address how to achieve scalable content distributions. We present two contributions, each of which uses a different approach to distribute the content.
In the first part of this chapter, we consider a terrestrial overlay network and build on top of it a VoD service for fixed clients. The goal is to minimize the operational cost of the service. Our contibutions are as follows. First, we introduce a new video distribution architecture that combines open-loop and closed-loop schemes. This combination makes the overall system highly scalable, very cost-effective, and ensures a zero start-up delay. Our second contribution is a mathematical model for the cost of delivering a video as a function of the popularity of that video. Our analytical model, along with some extensions, allows us to explore several scenarios: (i) long videos of 90 min (movies), (ii) short videos of a few min (clips), (iii) the dimensioning of a video on demand service from scratch, and (iv) the case of the optimization of an already installed video on demand service (i.e. the limited resources scenario).
In the second part of this chapter, we consider a satellite distribution of contents to mobile users, or in general to users thar are occasionally connected. We consider a push-based model, where the server periodically downloads objects. We assume that clients register to the service off-line. Our goal is to minimize the mean aggregate reception delay over all objects where each object is weighted by its popularity. Our contibutions in this part are as follows. First, we provide a simple proof for the need of periodicity (equal distance in transmission) of popular objects in a cycle. Second, in contrast to existing results, we consider the scheduling problem for caching clients. To increase the performance of the system, we also evaluate a pre-emptive scheduling algorithm that allows interruption (pre-emption) of an object’s transmission in order to transmit on schedule another more popular one.