We present a decentralized geospatial Web services composition method based on the peer-to-peer models, which partitions the global process model into different sub-process models on the corresponding peers. An algorithm for generating sub-process models is proposed to keep the control dependences and data dependences of the global process model. We also present a runtime load balancing algorithm to dynamically adjust the neighbor peers' sub-processes according to peer's load. The simulation results show that our approach can do well in the decentralized geospatial services composition with highly concurrency and large data volume as well as unbalanced peer-loads.
展开▼