Adjusting middleware knobs to assess scalability limits of distributed cyber-physical systems

Abstract

The traditional development paradigm for time-sensitive distributed systems (and even more for real-time domains) has typically relied on unflexible low-level schemes; these have been based on (also) low-level programming of improved medium access control protocols to obtain deterministic network schedules. Such technique does not scale well in the context of cyber-physical systems (CPS), which have a complexity of several orders of magnitude higher. In this paper, we explore the actual trend of cyber-physical systems (CPS) that are progressively integrated with Internet technologies to fulfill their requirements of being highly connected systems. The integration pillar is the communication middleware; however, communication middleware technologies are prone to introducing delays at different levels, increasing the potential uncertainty of the system execution. Therefore, individual analysis of middleware technologies is needed to assess the bounds on the type and scale of CPS that they can support. We analyse the behavior of a specific a middleware technology and its cost for handling the interaction of distributed nodes in the context of cyber-physical systems. We propose the design of a reliable middleware infrastructure for distributed embedded systems integrated in a CPS environment. Our approach considers time requirements specified by the system nodes or units, and fine tunes a few specific parameters of the middleware that we refer to as knobs. We validate our solution implementation in a experimental setting and show the system scale that can be supported in a stable way.