DECA: A Hierarchical Framework for DECentralized Aggregation in DHTs

As Structured Peer-to-Peer (P2P) Networks become popular, there is an emerging need to monitor continuously the huge number of participants in a robust and scalable manner. To this end, aggregation has emerged as a basis for the self-management of these networks. However, the structured P2P networks lack today of efficient mechanisms for the decentralized computation of these aggregates. In this paper, we propose a hierarchical theoretic model based on Cayley Graphs, which overcomes the requisite to accommodate growth without impacting the efficiency of distributed applications. Also, the paper presents an aggregation protocol that fuses the fault-resilience of gossip algorithms with the scalability of trees. In particular, simulation results show that this algorithm is capable to cope with the distributed and unreliable nature of P2P networks.