Research Project: Multipeer information dissemination and consistency support

Project description

This project is to study multi-peer info-dissemination and consistency support for collaborative environments and distributed objects suitable for such environments, together with efficient data structures and algorithms for their realization. Application domains which are of interest are telecommunication systems and distributed infrastructures for collaborative activities (cf also the group's related projects on resource allocation and sharing in network communication and collaborative environments and visualization. The main goal is to produce and make available new knowledge within this area, to offer improved capabilities for serving the society's needs in a cost-and-manner-effective way.

For the efficient and scalable dissemination of information in large systems, probabilistic dissemination protocols have appeared as an interesting alternative to deterministic network-level protocols. With probabilistic dissemination protocols, every participant periodically exchanges information with its neighbours. Such protocols hence are well-suited for peer-to-peer communication structures, since all participants (e.g., users of a collaborative environment) play the same role. Besides, they can help to achieve a high degree of reliability, i.e., every participant can potentially request or handle the retransmission of information. These properties and potential make probabilistic dissemination protocols especially interesting.

Recent results include the study of scalable logical clocks for ordering of events in distributed systems in a way that can offer significant savings in time overhead and transmission costs and the formulation of new, adaptive and scalable methods for such clocks, with improved accuracy, which is important in peer-to-peer communication systems.

Results achieved within the project include: distributed protocols for scalable information dissemination with efficient resource utilization; the study of scalable logical clocks for ordering of events in distributed systems in a way that can offer significant savings in time overhead and transmission costs; the formulation of new, adaptive and scalable methods for such clocks, with improved accuracy, which is expected to have significant impact in multi-peer communication systems.

The project is further going to build on these results in order tackle the issue of designing appropriate algorithms to handle filtering of information according to the individual interests and consistency requirements of the collaborative systems peers, i.e. the views that are of interest to the parts running the protocol.

Top of Page

Publications

• Anders Gidenstam, Boris Koldehofe, Marina Papatriantafilou and Philippas Tsigas,
  Dynamic and fault-tolerant cluster management,
  Proceedings of the 5th IEEE International Conference on Peer-to-Peer Computing, pages 237 - 244, IEEE Press, 2005.
• Anders Gidenstam, Boris Koldehofe, Marina Papatriantafilou and Philippas Tsigas,
  Lightweight Causal Cluster Consistency,
  Proceedings of the Conference on Innovative Internet Community Systems (I²CS 2005), LNCS, Springer Verlag, (To appear).
• Anders Gidenstam, Marina Papatriantafilou ``Adaptive Plausible Clocks'' Proceedings of the 24th International Conference on Distributed Computing Systems (ICDCS 2004), IEEE Press, 2004
• Boris Koldehofe ``Buffer Management in Probabilistic Peer-to-Peer Communication Protocols'' Proceedings of the 22nd Symposium on Reliable Distributed Systems (SRDS '03), October 2003, IEEE.
• Boris Koldehofe ``Simple gossiping with balls and bins'' Proceedings of the 6th Annual 6th International Conference on Principles of Distributed Systems(OPODIS '02), pages 109--117, December 2002.

Technical Reports

• Anders Gidenstam, Boris Koldehofe, Marina Papatriantafilou, Philippas Tsigas ``Dynamic and fault-tolerant cluster management'' Technical Report 2005-10, Computer Science and Engineering, Chalmers University of technology, 2005.
• Anders Gidenstam, Boris Koldehofe, Marina Papatriantafilou, Philippas Tsigas ``Lightweight Causal Cluster Consistency'' Technical Report 2005-09, Computer Science and Engineering, Chalmers University of technology, 2005. (This is an updated version of 2004-07 below.)
• Anders Gidenstam, Boris Koldehofe, Marina Papatriantafilou, Philippas Tsigas ``Lightweight Causal Cluster Consistency'' Technical Report 2004-07, Computing Science, Chalmers University of technology, 2004.

Top of Page