Reasoning of real-time distributed programming languages

Abstract

Two important features of a real-time distributed programs are: establishing the logical correctness of the program using realistic models and establishing timing properties or deriving timing constraints. In this paper, we provide a hierarchy of proof techniques for deriving the timing properties of real-time distributed programs. First, we provide a static characterization of real-time distributed programs using the maximum-parallelism model and show that several interesting timing characteristics can be derived from such an analysis. The static analysis is based on complete trace structures and hence, we can specify various classes of safety (including deadlock), evcutuality propcrtics uch as eventual stability (pcrsistance), recurrence, progress etc as well as parallel actions, and timing constraints. Based on prefix-closed tract structures (under the maximal parallelism model), we define two proof systems referred to as syntactic and general proof systems. The syntactic proof system uses assertions (postulates) attached to the I/O commands (essentially, reflecting the cnablcment, or othcrwisc of the guards) rather than state information. Using such a system, WC establish that we can refine several of the timing properties obtained by the static analysis. All the proof syntcms arc compositional and they form a hierarchy. The hierarchy not only establishes the consistency of the systems but also provides the user a selection of proof systems depending on the requirements. Further, the algorithms based on the static analysis lead to algorithmic derivation of the timing properties of a wide variety of real-time programs.