Novel Search Schemes for Distributed Cooperative Data Centers

Distributed cooperative data centers provide a new data storage and data processing architecture over heterogeneous physical nodes. The major challenge faced by such architecture is to find the right data on some node efficiently. In this paper, we present two alternatives for data indexing based on Chord, a Peer to Peer (P2P) overlay structure, which redefine the routing table structure that underlies the Chord and achieve small search paths. First, we present a regional search algorithm that routes data keys queries by region and super-node information. The search process can be easily implemented via O(log K) hops, while maintaining O(log K) routing information with K regions in the Chord ring at each node. We further propose a Two-hop search based on the regional search scheme which aims to reduce the average search paths to a constant with O(log K) routing states about super-nodes. Results from theoretical analysis and simulations show that our improved routing algorithms can achieve higher search efficiency and the improved membership maintenances can keep routing information sufficiently up-to-date to validate higher search successful rate.