Reducing maintenance overhead via heterogeneity in Skip Graphs

Skip Graphs provide novel distributed data structures for peer-to-peer (P2P) searching. They are simple, resilient and are capable of supporting complex queries. Like many distributed hash table (DHT) algorithms, one issue that remains to be addressed is the large number of pointers per node in the system. They suffer from the problems of high storage requirements and high volume of repair-mechanism message traffic. In addition, nodes are treated equally in Skip Graphs, but empirical studies have shown that peers in P2P networks are heterogeneous. Some nodes are more powerful and stable than the others. In order to solve these problems, we propose the Skip Graphs++, which gives an adaptive node capacity model to evaluate the performance of nodes in the system and builds the relationship between the loads and the capacities of nodes. In Skip Graphs++ system the loads are proportional to the capacities of nodes. Powerful nodes afford more loads and weak nodes afford fewer loads. Theoretical and experimental results have shown that the maintenance overhead is reduced efficiently via heterogeneity.