P2P scientific collaboration: structured topology and replication

摘要: The intrinsic instability of many unstructured peer-to-peer (P2P) networks due to continues nodes arrivals and departures can generate low availability of resources and a great overload due to searches for these resources. Our hypothesis is that these problems can be reduced if P2P application replicate content, and the relationships of the collaboration and common interests expressed by social networks are used for the interconnection of peers, and for selecting the location of objects. In this paper we present a study of replication mechanisms in P2P Scientific Collaboration Networks (SCN), which take advantage of the properties inherent in these networks (small-world, community structure), in order to select the location of objects, reduce the search costs, and increase the probability of reaching the target object. This affects both the performance of the applications and the efficient use of resources. The cost to retrieve an object depends on the diameter of the p2p network and the number and location of replica. As a result the search cost is dramatically reduced going from the diameter of the global network to the diameter of a cluster, and replicas are more effective as they are located close to the demand.