GrouPeer: Dynamic clustering of P2P databases

Abstract

Abstract. Sharing structured data in a P2P network is a challenging problem, especially in the absence of a mediated schema. The standard practice of answering a consecutively rewritten query along the propagation path often results in significant loss of information. On the opposite, the use of mediated schemas requires human interaction and global agreement, both during creation and maintenance. In this paper we present GrouPeer, an adaptive, automated approach to both issues in the context of unstructured P2P database overlays. By allowing peers to individually choose which rewritten version of a query to answer and evaluate the received answers, information-rich sources left hidden otherwise are discovered. Gradually, the overlay is restructured as semantically similar peers are clustered together. Experimental results show that our technique produces very accurate answers and builds clusters that are very close to the optimal ones by contacting a very small number of nodes in the overlay.

Semantic Grouping of Social Networks in P2P Database Settings

“…This can be achieved either manually or using one of the proposed schemes (e.g., [12,16]). Finally, we also assume that peer mappings between acquaintees are of the widely-known GAV/LAV/GLAV form [6,13] and peer schemas are relational, (i.e., the only internal mappings are foreign key constraints).…”

“…The latter is supposed to merge its own schema with the group schema it receives according to the merging procedure described in the section 2.2. Every peer P 1 For more details on concepts, see [12] V on the network path of the inference process determines its acquaintees P j for which SS(S I , S P j ) ≥ t, adds the respective pair P j , SS(S I , S P j ), to ST (I) and orders it. Any peer P on the inference process path calculates SS(S I , S P j ) indirectly, as the product: SS(S I , S P ) · SS(S ′ P , S P j ), where S ′ P is the part of S P mapped on S I .…”

“…Finally, we require that merging is only based on available information on the peers, i.e., it solely exploits the peer schemas and the peer mappings. Each mapping is considered to be a set of 1-1 correspondences between attributes that hold with an optional set of value and join constraints on some attributes (see [12] for details).…”

“…To evaluate the performance of the proposed group inference procedure, we use a message-level simulator that implements it over an unstructured overlay of semanti- The clustering is performed using the GrouPeer system [12]. In GrouPeer, peers decide to add new (and abolish old) one-hop neighbors in the overlay (acquaintees) according to the accuracy of the answers they receive from remote peers.…”

“…Based on these mappings, they decide to become acquainted with peers that store information similar to their interests. The result is an effective semantic clustering of the overlay, where the accuracy of query rewritings and answers is a lot higher compared to the unclustered overlay (for details see [12]). We compare the query evaluation performed by GrouPeer with the evaluation that utilizes the inferred groups on the overlay.…”