一组编制模式下组合Web服务模型的约简规则与算法

A Set of Model Reduction Rules and Algorithms for Web Services Composition under Orchestration Pattern

为了解决Web服务组合过程中状态空间庞大而导致性能求解的效率难以满足实时性要求的问题,提出了一种过程约简算法.在利用广义随机Petri网进行过程建模后,保留系统时间性能特征,通过对可约简子网的自动探察,剔除部分规则结构(序列、选择、并行、循环),孤立出可单独求解的子结构,由此在比较满意的时间内完成大规模模型约简.多种模型规模下的实验性能表明,所提算法可在动态服务组合与组合过程中自适应、快速地获得系统响应时间和吞吐量等性能指标,适用于多数服务组合过程的在线响应时间分析.

The performance calculation of Web services composition process in large state space is time-consuming, and is difficult to fulfill the real-time performance analysis. Therefore, a process reduction algorithm is presented to accelerate the calculation of system performance. The Web services composition process is modeled using the generalized stochastic Petri net. Through autodetecting the reductive subnet, some regular structures, such as sequence, choice, concurrent and loop structures, can be eliminated under the precondition of preserving the timing constraints. The sub processes which can be calculated independently are separated. A large scale model can thus be reduced within a certain period. Experimental results with multiform models indicate that the system response time and throughput are obtained rapidly by using the algorithm in the dynamic and adaptive composition process. The algorithm can be applied to online performance analysis of most Web services composition process.