中文摘要：

空间服务组合执行过程中需要处理复杂多样的空间数据类型和大量的空间数据,采用集中式组合模型完成空间服务组合的通信开销代价将严重影响组合系统性能。分散式组合模型能有效克服上述缺点,并具备可扩性好,吞吐率高等其他特点。提出基于P2P网络的空间服务组合执行技术,通过Geo-Peer来形成P2P重叠网络从而构成空间服务组合分散式执行的基础设施。在执行时,首先将用户建模的流程图转化为能最大化表达并行度的最优流程图,其后将最优流程图分解为可部署到Geo-Peer上的多个子流程图,为全局流程分散到Geo-Peer网络的执行制定流程语义约束。模拟实验表明,基于P2P网络的空间服务组合执行能克服集中式组合的诸多局限,有效提升组合系统的整体性能。

英文摘要：

The centralized execution of service composition spends so much time in exchanging message that it will badly break down the performance of composition system. Hence, it can＇ t meet the requirement of geospatial service composition which needs to deal with complex spatial data types and enormous spatial data. Then, we present a novel execution model based on the Peer-to-Peer net- works, namely, decentralized execution model ,which has more powerful sealabihty and throughput, to implement geospatial service composition. In our model, the Geo-Peers build the infrastructures of P2P network to support decentralized composition. At the runtime, we first convert the original process graph into a Best Process Graph with the goal of maximizing the parallelisms between process activities, and then partition the Best Process Graph into several sub process graphs. All the sub process graphs can be directly deployed into different Geo-Peers and will help these Geo-Peers to coordinate with each other under the constraints of global process semantic. The experiment results show that the decentralized execution of geospatial service composition based on P2P network can conquer many problems which exist in centralized execution. So,it is helpful to improve the whole composition system＇ s performance.