Knowledge Based Transactional Behavior

Component-based systems (including distributed programs and multiagent systems) involve a lot of coordination. This coordination is done in the background, and is transparent to the operation of the system. The reason for this overhead is the interplay between concurrency and non-deterministic choice: processes alternate between progressing independently and coordinating with other processes, where coordination can involve multiple choices of the participating components. This kind of interactions appeared as early as some of the main communication-based programming languages, where overhead effort often causes a restriction on the possible coordination. With the goal of enhancing the efficiency of coordination for component-based systems, we propose here a method for coordination-based on the precalculation of the knowledge of processes and coordination agents. This knowledge can be used to lift part of the communication or synchronization that appears in the background of the execution to support the interaction. Our knowledge-based method is orthogonal to the actual algorithms or primitives that are used to guarantee the synchronization: it only removes messages conveying information that knowledge can infer.