Skip to Main Content
The distributed optimal path planning (D-OPP) problem has been a bottleneck restricting the performance of Peer-to-Peer (P2P) workflow systems which have met the development trend of workflow systems and been one of the most important domains of Distributed computing. The essence of this problem is to solve the dynamical global optimization in local views without centralized mechanism. Aiming at this problem, an innovative hybrid P2P location network consisting of spanning graph location network (SGLN) and workload balancing network (WBN) is proposed based on a novel variable length coding which encodes the logical relationship of workflow model. SGLN achieves one hop routing complexity with 100% location precision and zero extra routing bandwidth at a coarse granularity. And WBN realizes an optimal decentralized dynamic task scheduling at a fine granularity by constructing a randomized network which can keep stable topology and a heuristic random sampling with fast converge at the optimal peers. The combination of SGLN and WBN provides an optimal solution to the D-OPP problem and a novel cooperative computing paradigm to distributed computing.
Date of Conference: 8-11 Dec. 2010