基于数据分发服务的远程过程调用系统

Remote procedure call system based on data distribution service

随着分布式系统的发展,用户对于数据传输的需求变得多样化,在大型分布式系统中通常需要不同的通信方式。然而目前大部分中间件仅支持单一通信方式,部署多个中间件则需占用较多资源。针对这项需求,在发布订阅中间件数据分发服务(DDS)上实现使用请求应答通信方式的远程过程调用系统(RPC Over DDS),使用户在分布式系统中部署一个中间件即可同时使用发布订阅和请求应答两种通信方式,并设计了两种服务端线程模型以满足用户在不同场景下的需求。此外,还对服务端两种线程模型及DDS和RPC Over DDS的性能差异进行了比较,结果表明服务端两种线程模型适用于不同的应用场景,而DDS与RPC Over DDS的性能差距仅在5. 5%左右,能支持快速高效的点对点数据传输。

With the development of distributed systems,the demands of data transmission have been diversified,so different communication styles are generally required in large-scale distributed systems.However,most middlewares currently support only a single communication style,while it requires more resources to deploy multiple middlewares.Relied on this requirement,the authors implemented a Remote Procedure Call(RPC)system using request-reply communication style based on Data Distribution Service(DDS)using publish-subscribe communication style(RPC Over DDS).It enables users to use both publish-subscribe communication style and request-reply communication style by deploying only one middleware in a distributed system.Besides this,the authors designed two server thread models to satisfy the needs in different application scenarios.In addition,the authors also made a comparison between the two server thread models and analyzed the performance differences between DDS and RPC Over DDS.The results show that the performance of the two server thread models is different in different application scenarios.While the performance gap between DDS and RPC Over DDS is about 5.5%,it proves that RPC Over DDS is able to support fast and efficient point-to-point data transmission.