Peer-based programming model for coordination patterns

Abstract

Modern distributed software systems must integrate in near-time parallel processes and heterogeneous information sources provided by active, autonomous software systems. Such lively information sources are e.g. sensory data, weather data, traffic data, or booking data, operated by independent distributed sites. The complex integration requires the coordination of these data flows to guarantee consistent global semantics. Design, implementation, analysis and control of distributed concurrent systems are notoriously complex tasks. Petri Nets are widely used to model concurrent activities. However, a higher-level programming abstraction is needed. We propose a new programming model for modeling concurrent coordination patterns, which is based on the idea of "peer workers" that represent re-usable coordination and application components. These components encapsulate behavior, structure distributed data and control flow, and allow integration of pre-existing service functions. A domain-specific language is presented. The usability of the peer-based programming model is evaluated with the Split/Join pattern. © 2013 IFIP International Federation for Information Processing.