基于SVM机器学习的仿真网格资源调度模型

摘要

According to the characteristics of simulation grid resources (SGR),an extended Web Service Description Language (WSDL) was used to describe the attributes of the SGRs,which is conducive to the application of machine learning algorithms for SGR resource scheduling on a centralized-distributed SGR management model.These specific requirements of Distributed Interactive Simulation (DIS) based on the task were analyzed;and SVM-based machine learning resource scheduling simulation model was proposed.Support Vector Machine (SVM) and incremental SVM were applied to implement SGRs classification when the features vectors were extracted from the WSDL documents.Scheduling agents can then carried out the SGR scheduling on classified SGRs.Experimental results show that the scheduling model can get Federation overall optimal result with better performance.%根据仿真网格资源的特性,采用扩展的WSDL描述仿真网格资源的属性,应用机器学习算法进行资源调度,建立了集中-分布式仿真资源管理模型；在分析分布交互仿真任务特有需求的基础上,提出了一种基于支持向量机(support vector machine,SVM)机器学习的仿真资源调度模型.该模型基于资源描述提取资源的特征量进行规范化处理,采用SVM和增量SVM技术进行分类,调度代理在分类训练结果的基础上进行调度.实验表明,该调度模型可以更好的性能获得联邦(Federation)整体最优的调度结果.