Virtualizing DDS middleware: Performance challenges and measurements

Abstract

As new technology becomes available, systems increase in complexity, which in turn, raises the expectations of users for new applications that may presumably be more complex. This spiral process requires the usage of appropriate techniques to control complexity such as decoupled design and development paradigms, and communications middleware that facilitate the development of distributed applications. The highest exponent of them is, currently, DDS middleware (Data Distribution System for Distributed Real-Time Systems) that is specifically designed for applications that have timing requirements. Also, virtualization techniques follow the principle of complexity reduction, and they allow to customize the offered computational platforms and to achieve server consolidation in different domains ranging from industrial control systems, distributed surveillance, or enterprise resource planning applications. This paper describes some considerations for merging real-time middleware, such as DDS, and virtualization technology with the aim of suiting the cyber physical domain. This virtual integration tries to bypass the typical bottlenecks of performance. Performance results track the differences between the executions of DDS middleware in the bare machine compared to a virtualized environment.