Distributed computer architectures labeled “peer-to-peer ” are designed for the sharing of computer resources (content, storage, CPU cycles) by direct exchange, rather than requiring the intermediation or support of a centralized server or authority. Peer-to-peer architectures are characterized by their ability to adapt to failures and

Recently, peer-to-peer (P2P) networks have emerged as an attractive solution to enable large-scale content distribution without requiring major infrastructure investments. While such P2P solutions appear highly beneficial for content providers and end-users, there seems to be a growing concern among Internet Service Providers (ISPs) that now need to support the distribution cost. In this work, we explore the potential impact of future P2P file delivery mechanisms as seen from three different perspectives: i) the content provider, ii) the ISPs, and iii) individual content consumers. Using a diverse set of measurements including BitTorrent tracker logs and payload packet traces collected at the edge of a 20,000 user access network, we quantify the impact of peer-assisted file delivery on end-user experience and resource consumption. We further compare it with the performance expected from traditional distribution mechanisms based on large server farms and Content Distribution Networks (CDNs).

Available grid technologies like the Globus Toolkit make possible for one to run a parallel application on resources distributed across several administrative domains. Most grid computing users, however, don't have access to more than a handful of resources onto which they can use this technologies. This happens mainly because gaining access to resources still depends on personal negotiations between the user and each resource owner. To address this problem, we are developing the OurGrid resources sharing system, a peer-to-peer network of sites that share resources equitably in order to form a grid to which they all have access. The resources are shared accordingly to a network of favors model, in which each peer prioritizes those who have credit in their past history of bilateral interactions. The emergent behavior in the system is that peers that contribute more to the community are prioritized when they request resources. We expect, with OurGrid, to solve the access gaining problem for users of bag-of-tasks applications (those parallel applications whose tasks are independent).

Motivated by the need to deploy a public repository of multi-gigabyte trace files, we studied the BitTorrent protocol’s ability to disseminate very large files among peers. BitTorrent is a popular peer-to-peer protocol that allows parallel downloads of large files. In this paper, we analyzed user activity on BitTorrent over a four-month period with respect to supportable file sizes, file popularity, session lengths, transfer speeds, and the likelihood of service-interrupting flash crowds. Our results show that file sizes tend to be on the order of gigabytes, far larger than other peer-to-peer applications. File popularity has a distribution similar to other peer-to-peer file sharing systems. Unlike other systems, the majority of users require multiple sessions to retrieve a file, and they are willing to remain connected to the system for a very long time. Most users we observed appear to have asymmetric Internet connections, and their generally poor upload performance is mitigated by their willingness to remain connected to the system and upload for an amount of time far longer than they spent downloading. We found that service disruption due to flash crowds is unlikely, as the vast majority of users were able to begin contributing resources back to the system within seconds of connecting. Our results indicate that BitTorrent provides an effective foundation for dissemination of files that are multi-gigabyte or larger, provided more sophisticated features are added like versioning, availability, and content management. I.

Peer-to-peer (P2P) systems are typically divided into those that centralize lookup functionality in a single location and those that distribute the lookup operation across the set of participating hosts. The former approach can offer constant time lookup latency, but is more expensive to scale and suffers from single points of failure. In contrast, the fully distributed approach is easier to scale and can be more resilient to failures, but the lookup latency scales as a function of the total number of participants. While the research community has made great progress in improving the latency of distributed lookup, these systems, exemplified by Chord[17], typically require O(logN) hops to locate an object in a system with N hosts. In this paper, we explore the costs and benefits of a new hybrid approach that partially distributes lookup information among a dynamically adjusted set of high-capacity "superpeers". This design exploits the resource heterogeneity inherent in existing P2P systems to provide many of the advantages of a centralized system, even while avoiding most of the problems associated with such systems. Lookup is performed using superpeers in constant-time, and the system performs well even in the event of simultaneous superpeer failures. Finally, while our gain in performance is potentially at the expense of scalability, we will show that a straightforward implementation should be able to scale to over one million peers with reasonable lookup rates.

Assessing the performance of peer-to-peer algorithms such as topology construction protocols, distributed trust or search algorithms is impossible without simulations since testing new algorithms by deploying them in an existing P2P network is prohibitively expensive. However, some P2P algorithms are sensitive to the network and traffic models that are used in the simulations. In order to produce realistic results, we therefore require models that resemble real-world P2P networks as closely as possible. In this paper, we describe a model for P2P file-sharing networks, link it to measurements on existing P2P networks and discuss open issues in modeling these networks.

Cooperation is a central tenet of peer-to-peer systems. Without cooperation, users of a file-sharing system such as Gnutella suffer long download delays, if they are able to download at all [1]. Unfortunately, users have a natural disincentive

Grid computing has excited many with the promise of access to huge amounts of resources distributed across the globe. However, there are no largely adopted solutions for automatically assembling grids, and this limits the scale of today's grids. Some argue that this is due to the overwhelming complexity of the proposed economy-based solutions. Peer-to-peer grids have emerged as a less complex alternative. We are currently deploying OurGrid, one such peerto -peer grid. OurGrid is a CPU-sharing grid that targets Bag-of-Tasks applications (i.e. parallel applications whose tasks are independent). In order to ease system deployment, OurGrid is based on a very lightweight autonomous reputation scheme.

In the recent years, the evolution of a new wave of innovative network architectures labeled peer-to-peer (p2p) has been witnessed. Such architectures and systems are characterized by direct access between peer computers, rather than through a centralized server. P2p file sharing architectures can be classified by their degree of centralization , i.e. to what extent they rely to one or more servers to facilitate the interaction between peers. Three categories are identified: Purely decentralized, partially centralized and hybrid decentralized. Furthermore, highly dynamic p2p networks of peers with complex topology can be differentiated by the degree to which they contain some structure or are created adhoc. By structure we refer to the way in which the content of the network is located: Is there a way of directly knowing which peers contain some specific content, or does one need to randomly search the entire network to locate it? Three categories of systems are examined: Structured, loosely structured and unstructured. Various p2p architectures from these categories are examined with focus on the way they operate and how successfully they address issues such as scalability, network latency, security, privacy, anonymity and others. The shortcomings of these systems and the latest variations, developments and trends in p2p file sharing network design that aim at improving upon them, are discussed.

Peer-to-Peer (P2P) systems have become popular over the past few years. However, their large scale and the open nature of the system makes studying them challenging. This paper presents an extensible framework for simulating P2P networks efficiently and accurately. Efficiency is accomplished by using message level simulation rather than packet level simulation. Moreover, accuracy is maintained by tracking the network infrastructure and using a flow model to accomplish accurate estimate of the message behavior. A second contribution of the paper is to model the BitTorrent (BT) protocol. BT is a widely-used protocol that is significantly more complex than other P2P protocols because file download occurs in chunks from many other peers concurrently. Thus, contrary to models of other P2P systems such as Gnutella or Freenet, which focus on finding the location of a file in the network, BT’s complexity occurs in downloading files (locating files in fact occurs out of band using websites that host the Torrent files). We validate the model against a packet level simulator and also using a real, but small scale, BitTorrent experiment. The simulator is object oriented and extensible for simulating other P2P protocols and applications. 1