We are building Coral, a peer-to-peer content distribution system. Coral creates self-organizing clusters of nodes that fetch information from each other to avoid communicating with more distant or heavily-loaded servers. Coral indexes data, but does not store it. The actual content resides where it is used, such as in nodes' local web caches. Thus, replication happens exactly in proportion to demand.

DS is a dependable and secure data storage for mobile, wireless networks based on a peer-to-peer paradigm. DS share files under a write-once model, and ensures data confidentiality and dependability by encoding files in a Redundant Residue Number System. The paper analyzes the code efficiency of DS using a set of moduli allowing for efficient encoding and decoding procedures based on single precision arithmetic, with the Information Dispersal Algorithm approach (IDA) shows that DS features security features which are not provided by IDA, while the two approaches are comparable from the viewpoint of code efficiency and encoding/decoding complexity.

In this paper we focus on the design of high performance peer-to-peer content sharing systems. In particular, our goal is to achieve global load balancing and short user-request response times. This is a formidable challenge, given the requirement to respect the autonomy of peers, their heterogeneity in terms of processing and storage capacities, their different content contributions, the huge system scale, and the dynamic system environment. Our approach exploits the semantic categorization of published documents and constructs clusters of peers. We provide a formal formulation for the problem of load balancing in our setting and prove that it is NP-complete. We also present a greedy polynomial time algorithm that achieves nearly optimal load balancing as shown by our experimental results.

to advance, and has reached a point where the potential benefits of very large scale, finely distributed applications are more apparent than ever. Opportunities are emerging to develop large systems that cater to highly dynamic and mobile sets of participants, who desire to interact with each other and stores of online content in a robust manner. These opportunities will inevitably dictate a substantial body of research in the years to follow. Although applications intended to function at this scale have recently begun to appear, there remain a broad set of open problems that must be faced before this emerging class of distributed system can become a reality. 1

We are building Coral, a peer-to-peer content distribution system. Coral creates self-organizing clusters of nodes that fetch information from each other to avoid communicating with more distant or heavily-loaded servers. Coral indexes data, but does not store it. The actual content resides where it is used, such as in nodes' local web caches. Thus, replication happens exactly in proportion to demand.

this paper, we use the word "repair" rather than "recover" to emphasize the additional, often manual, efforts required to preserve useful data in the process of restoring the system back to normal order

References are ubiquitous in computing. Memory addresses, URLs, file names are references. Every reference has a referent - the data that is referred to. By the reference-referent relation we have in mind the relation between two pieces of data, where given the first piece of data (reference) it is possible to retrieve the other piece (referent) unambiguously, and in a way that was intended by the designer of the system. Commonly, the relationship between reference and referent is defined relative to the state of some physical system. Memory addresses refer to the contents of a particular section of physical memory on a particular machine. URLs refer to the contents of a file on a given web server. File names refer to the contents of a particular section of a physical disk. Consequently, if the state of the physical system changes, the referent changes too. We call these types of references, where the relationship between the reference and referent is defined by the state of a phy

We describe the implementation and evaluate the performance of a Replica Location Service that is part of the Globus Toolkit Version 3.0. A Replica Location Service (RLS) provides a mechanism for registering the existence of replicas and discovering them. Features of our implementation include the use of soft state update protocols to populate a distributed index and optional Bloom filter compression to reduce the size of these updates. Our results demonstrate that RLS performance scales well for individual servers with millions of entries and up to 100 requesting threads. We also show that the distributed RLS index scales well when using Bloom filter compression for wide area updates.