Dynamic load balancing for distributed memory multiprocessors
Journal of Parallel and Distributed Computing
Load balancing and Poisson equation in a graph
Concurrency: Practice and Experience
Randomized algorithms
Freenet: a distributed anonymous information storage and retrieval system
International workshop on Designing privacy enhancing technologies: design issues in anonymity and unobservability
Search and replication in unstructured peer-to-peer networks
ICS '02 Proceedings of the 16th international conference on Supercomputing
IEEE Internet Computing
Diffusive Load-Balancing Policies for Dynamic Applications
IEEE Concurrency
Replication strategies in unstructured peer-to-peer networks
Proceedings of the 2002 conference on Applications, technologies, architectures, and protocols for computer communications
Pastry: Scalable, Decentralized Object Location, and Routing for Large-Scale Peer-to-Peer Systems
Middleware '01 Proceedings of the IFIP/ACM International Conference on Distributed Systems Platforms Heidelberg
Simple efficient load balancing algorithms for peer-to-peer systems
Proceedings of the sixteenth annual ACM symposium on Parallelism in algorithms and architectures
A survey and comparison of peer-to-peer overlay network schemes
IEEE Communications Surveys & Tutorials
Performance evaluation of replication strategies in DHTs under churn
Proceedings of the 6th international conference on Mobile and ubiquitous multimedia
Hi-index | 0.00 |
The present paper introduces a replication method that is meant to balance the storage load of peers in unstructured peer-to-peer (P2P) networks for file sharing and to provide good search performance. According to the random walk theory on an arbitrary network, the frequency of arrival of a random walker to a peer is proportional to the degree of the peers. Therefore, to limit the increase in the number of hops required to find a requested file, it is better to make as many files as possible in peers of high degree when using random-walk-based query forwarding methods. However, this causes a load bias to peers of high degree. That is, there is a trade-off between storage load balancing and search performance. The replication method presented herein replicates a requested file in a peer of low load adjacent to a peer of interest on the present search path by using dynamically varying values that represent the state of the load of peers. Therefore, it is expected that local storage load balancing will be achieved. Furthermore, since the proposed method causes peers adjacent to peers of high degree to hold several files, it is also expected that good search performance will be obtained. In addition, a replication method that makes a replica of a requested file in a peer on the present search path with probability inversely proportional to the degree of the peer is prepared as a method for comparison. The probability is, in contrast to the proposed method, statically determined prior to the start of the search according to the theory to ensure good load balancing. The experimental results show that both the proposed method and the method for comparison achieve global load balancing, although they have totally different strategies in terms of storage load balancing. It is, however, shown that only the proposed method does not require appropriate adjustment of parameter values to a given network topology prior to the start of the search in order to achieve good search performance. This is a significant advantage over the method for comparison.