Coordination of massively concurrent activities

Massively parallel and distributed systems open new horizons for large applications and present new challenges for software technology. Many applications already take advantage of the increased raw computational power provided by such parallel systems to yield significantly shorter turn-around times. However, the availability of so many processors to work on a single application presents a new challenge to software technology: coordination of the cooperation of large numbers of concurrent active entities. Classical views of concurrency in programming languages that are based on extensions of the sequential programming paradigm are ill-suited to meet this challenge. Exploiting the full potential of massively parallel systems requires programming models that explicitly deal with the concurrency of cooperation among very large numbers of active entities that comprise a single application. In practice, the concurrent applications of today essentially use a set of ad hoc templates to coordinate the cooperation of their active components. This shows the lack of proper coordination languages that can be used to explicitly describe complex cooperation protocols in terms of simple primitives and structuring constructs. In this paper we present a generic model of communication and describe a specific control-oriented coordination language based on this model. The important characteristics of this model include compositionality, which it inherits from the data-flow model, anonymous communication, and separation of computation concerns from communication concerns. These characteristics lead to clear advantages in large concurrent applications.