Unstructured peer-to-peer (P2P) architectures offer several benefits to implement semantic discovery and composition in future-generation service registries. However, their success strongly depends on the adoption of efficient techniques for disseminating semantic queries over the network. Gossip strategies significantly reduce the amount of messages with respect to flooding, but they need a predefined tuning of the effectual fanout to achieve good performance. In this paper, we compare typical gossip strategies with our proposal, which is able to dynamically exploit network knowledge to fulfil a selective choice of propagation paths in order to ensure high recall and further reduce the number of messages exchanged. We perform the comparison in a simulated environment to observe resolution time, recall and message overhead on large-size and evolving networks while searching for service compositions. We have adopted Bernoulli, Random Geometric and Scale-Free graphs to model different network topologies. The experimental results show that our approach is able to adapt to network changes and preserve high levels of recall. In particular, it reduces message overhead, with respect to both optimized flooding and the analysed gossip-based strategies, or improves the recall, whereas resolution time remains almost unchanged.

Gossip Strategies for Service Composition

Zimeo E.
2014-01-01

Abstract

Unstructured peer-to-peer (P2P) architectures offer several benefits to implement semantic discovery and composition in future-generation service registries. However, their success strongly depends on the adoption of efficient techniques for disseminating semantic queries over the network. Gossip strategies significantly reduce the amount of messages with respect to flooding, but they need a predefined tuning of the effectual fanout to achieve good performance. In this paper, we compare typical gossip strategies with our proposal, which is able to dynamically exploit network knowledge to fulfil a selective choice of propagation paths in order to ensure high recall and further reduce the number of messages exchanged. We perform the comparison in a simulated environment to observe resolution time, recall and message overhead on large-size and evolving networks while searching for service compositions. We have adopted Bernoulli, Random Geometric and Scale-Free graphs to model different network topologies. The experimental results show that our approach is able to adapt to network changes and preserve high levels of recall. In particular, it reduces message overhead, with respect to both optimized flooding and the analysed gossip-based strategies, or improves the recall, whereas resolution time remains almost unchanged.
2014
978-1-4799-2728-9
Gossip algorithms, Peer-to-Peer Computing, Query Forwarding, Service Composition, Service Discovery
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.12070/13070
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 1
  • ???jsp.display-item.citation.isi??? 1
social impact