Why a Distributed Solution?
Kluwer Academic Publishers eBooks · 2006
From the functional point of view, it makes little difference whether a given specification is implemented using a centralized architecture or a distributed architecture. In this chapter, a number of arguments are presented in favor of a distributed approach for the implementation of a hard real-time system. The chapter starts with an overview of a distributed real-time system architecture, which consists of a set of nodes and a communication network that interconnects these nodes. The interface between the communication network and the host computer within a node is discussed at length, and the concept of event message and state message is introduced. The semantics of state messages facilitates the exchange of state information among the nodes, and enforces a high degree of autonomy of the nodes and the communication system. The next sections argue for a distributed architecture as the preferred alternative for the implementation of a hard real-time system. The first and most important argument is that of composability. In a composable architecture, system properties follow from subsystem properties. In such an architecture, a constructive approach to system building and to system validation is supported. In the context of real-time systems, composability requires that the communication network interface between the host computer in a node and the communication network is fully specified, not only in the value domain, but also in the time domain. Scalability requires that there exist no limits on the extensibility of a system, and that the complexity of reasoning about the proper operation of any system function be independent of the system size. It is shown that the economics of VLSI production favor the distributed approach for building a scalable large real-time system. Dependability arguments advocate a distributed approach because such an approach makes it possible to implement welldefined error-containment regions, and to achieve fault tolerance by replicating nodes.