基于RTI的集群计算中间件

以大运算量和大记录数据量为特征的大规模数学仿真系统，可以采用集群计算的方式实现。集群计算可分为基于单一操作系统和基于中间件两种模式。支持集群计算的中间件可以基于分布式对象实现，也可以基于HLA／RTI实现。该文分析了基于RTI的中间件（RTI Based Middleware，RBM）应能满足的基本要求，讨论了陆军师作战仿真系统中RBM的设计与实现问题。在该系统中，RBM被设计为一个接口类，全面封装了所用到的RTI服务，并为仿真应用程序提供了方便的访问接口。测试表明，RBM能够满足支持集群计算的大规模数据交互能力需求。

The RTI Based Middleware for Cluster Computing

Large-scale mathematical simulation systems,characterized by large computation and large amounts of record data,can be implemented by means of cluster computation.Cluster computation can be based on a single operation system,as well as a middleware.The middleware can be implemented via distributed objects or HLA/RTI.In this paper,some essential requirements that must be met by the RTI Based Middleware(RBM) are analyzed.Design and implementation of the RBM in the land division battle simulation system are discussed.The RBM is designed as an interface class that encapsulates all the RTI services used and offers some methods for facilitating access of the simulation application program.A test shows that RBM can meet the requirement of large amounts of interaction data in a cluster computation system.