This paper presents the design and evaluation of Pastry, a scalable, distributed object location and routing substrate for wide-area peer-to-peer applications. Pastry performs application-level routing and object location in a potentially very large overlay network of nodes connected via the Internet. It can be used to support a variety of peer-to-peer applications, including global data storage, data sharing, group communication and naming. Each node in the Pastry network has a unique identifier (nodeId). When presented with a message and a key, a Pastry node efficiently routes the message to the node with a nodeId that is numerically closest to the key, among all currently live Pastry nodes. Each Pastry node keeps track of its immediate neighbors in the nodeId space, and notifies applications of new node arrivals, node failures and recoveries. Pastry takes into account network locality; it seeks to minimize the distance messages travel, according to a to scalar proximity metric like the number of IP routing hops. Pastry is completely decentralized, scalable, and self-organizing; it automatically adapts to the arrival, departure and failure of nodes. Experimental results obtained with a prototype implementation on an emulated network of up to 100,000 nodes confirm Pastry’s scalability and efficiency, its ability to self-organize and adapt to node failures, and its good network locality properties.

The Cooperative File System (CFS) is a new peer-to-peer readonly storage system that provides provable guarantees for the efficiency, robustness, and load-balance of file storage and retrieval. CFS does this with a completely decentralized architecture that can scale to large systems. CFS servers provide a distributed hash table (DHash) for block storage. CFS clients interpret DHash blocks as a file system. DHash distributes and caches blocks at a fine granularity to achieve load balance, uses replication for robustness, and decreases latency with server selection. DHash finds blocks using the Chord location protocol, which operates in time logarithmic in the number of servers. CFS is implemented using the SFS file system toolkit and runs on Linux, OpenBSD, and FreeBSD. Experience on a globally deployed prototype shows that CFS delivers data to clients as fast as FTP. Controlled tests show that CFS is scalable: with 4,096 servers, looking up a block of data involves contacting only seven servers. The tests also demonstrate nearly perfect robustness and unimpaired performance even when as many as half the servers fail.

This paper argues that a new class of geographically distributed network services is emerging, and that the most effective way to design, evaluate, and deploy these services is by using an overlay-based testbed. Unlike conventional network testbeds, however, we advocate an approach that supports both researchers that want to develop new services, and clients that want to use them. This dual use, in turn, suggests four design principles that are not widely supported in existing testbeds: services should be able to run continuously and access a slice of the overlay’s resources, control over resources should be distributed, overlay management services should be unbundled and run in their own slices, and APIs should be designed to promote application development. We believe a testbed that supports these design principles will facilitate the emergence of a new serviceoriented network architecture. Towards this end, the paper also briefly describes PlanetLab, an overlay network being designed with these four principles in mind. 1.

Abstract Decentralized and unstructured peer-to-peer networks such as Gnutella are attractive for certain applicationsbecause they require no centralized directories and no precise control over network topologies and data placement. However, the flooding-based query algorithm used in Gnutella does not scale; each individual query gener-ates a large amount of traffic and, as it grows, the system quickly becomes overwhelmed with the query-induced load. This paper explores, through simulation, various alternatives to gnutella's query algorithm, data replicationmethod, and network topology. We propose a query algorithm based on multiple random walks that resolves queries almost as quickly as gnutella's flooding method while reducing the network traffic by two orders of mag-nitude in many cases. We also present a distributed replication strategy that yields close-to-optimal performance. Finally, we find that among the various network topologies we consider, uniform random graphs yield the bestperformance. 1 Introduction The computer science community has become accustomed to the Internet's continuing rapid growth, but even tosuch jaded observers the explosive increase in Peer-to-Peer (P2P) network usage has been astounding. Within a few months of Napster's [12] introduction in 1999 the system had spread widely, and recent measurement data suggeststhat P2P applications are having a very significant and rapidly growing impact on Internet traffic [11, 15]. Therefore, it is important to study the performance and scalability of these P2P networks. Currently, there are several different architectures for P2P networks:

This paper presents Scribe, a scalable application-level multicast infrastructure. Scribe supports large numbers of groups, with a potentially large number of members per group. Scribe is built on top of Pastry, a generic peer-to-peer object location and routing substrate overlayed on the Internet, and leverages Pastry's reliability, self-organization, and locality properties. Pastry is used to create and manage groups and to build efficient multicast trees for the dissemination of messages to each group. Scribe provides best-effort reliability guarantees, but we outline how an application can extend Scribe to provide stronger reliability. Simulation results, based on a realistic network topology model, show that Scribe scales across a wide range of groups and group sizes. Also, it balances the load on the nodes while achieving acceptable delay and link stress when compared to IP multicast.

We present Tapestry, a peer-to-peer overlay routing infrastructure offering efficient, scalable, locationindependent routing of messages directly to nearby copies of an object or service using only localized resources. Tapestry supports a generic Decentralized Object Location and Routing (DOLR) API using a self-repairing, softstate based routing layer. This paper presents the Tapestry architecture, algorithms, and implementation. It explores the behavior of a Tapestry deployment on PlanetLab, a global testbed of approximately 100 machines. Experimental results show that Tapestry exhibits stable behavior and performance as an overlay, despite the instability of the underlying network layers. Several widely-distributed applications have been implemented on Tapestry, illustrating its utility as a deployment infrastructure.

Structured peer-to-peer overlay networks provide a sub-strate for the construction of large-scale, decentralized applications, including distributed storage, group com-munication, and content distribution. These overlays are highly resilient; they can route messages correctly even when a large fraction of the nodes crash or the network partitions. But current overlays are not secure; even a small fraction of malicious nodes can prevent correct message delivery throughout the overlay. This prob-lem is particularly serious in open peer-to-peer systems, where many diverse, autonomous parties without pre-existing trust relationships wish to pool their resources. This paper studies attacks aimed at preventing correct message delivery in structured peer-to-peer overlays and presents defenses to these attacks. We describe and eval-uate techniques that allow nodes to join the overlay, to maintain routing state, and to forward messages securely in the presence of malicious nodes. 1

In this paper, we discuss the problem of distributing streaming media content, both live and on-demand, to a large number of hosts in a scalable way. Our work is set in the context of the traditional client-server framework. Specifically, we consider the problem that arises when the server is overwhelmed by the volume of requests from its clients. As a solution, we propose Cooperative Networking (CoopNet), where clients cooperate to distribute content, thereby alleviating the load on the server. We discuss the proposed solution in some detail, pointing out the interesting research issues that arise, and present a preliminary evaluation using traces gathered at a busy news site during the flash crowd that occurred on September 11, 2001.

Despite their growing popularity, the behavior of super-peernetworks is not well understood. For example, what are the potential drawbacks of super-peer networks? How can super-peers be made more reliable? How many clients should a superpeer take on to maximize efficiency? In this paper we examinesuper-peer networks in detail, gaining an understanding of their fundamental characteristics and performance tradeoffs. We alsopresent practical guidelines and a general procedure for the design of an efficient super-peer network.

Abstract. ABloomfilter is a simple space-efficient randomized data structure for representing a set in order to support membership queries. Bloom filters allow false positives but the space savings often outweigh this drawback when the probability of an error is controlled. Bloom filters have been used in database applications since the 1970s, but only in recent years have they become popular in the networking literature. The aim of this paper is to survey the ways in which Bloom filters have been used and modified in a variety of network problems, with the aim of providing a unified mathematical and practical framework for understanding them and stimulating their use in future applications. 1.