Replica-Based High-Performance Tuple Space Computing

We present the tuple-based coordination language RepliKlaim, which enriches Klaim with primitives for replica-aware coordination. Our overall goal is to offer suitable solutions to the challenging problems of data distribution and locality in large-scale high performance computing. In particular, RepliKlaim allows the programmer to specify and coordinate the replication of shared data items and the desired consistency properties. The programmer can hence exploit such flexible mechanisms to adapt data distribution and locality to the needs of the application, so to improve performance in terms of concurrency and data access. We investigate issues related to replica consistency, provide an operational semantics that guides the implementation of the language, and discuss the main synchronization mechanisms of our prototypical run-time framework. Finally, we provide a performance analysis, which includes scenarios where replica-based specifications and relaxed consistency provide significant performance gains.