Parallel computational model for multiprocessor systems with distributed memory

The emergence of powerful multiprocessor computing systems brings to the fore issues related to the development of frameworks (templates) that allow creating high-scalable parallel programs oriented to systems with distributed memory. In this case, the most important problem is the development of parallel computing models that allow us to assess its scalability at an early stage of the program design. General requirements for computational models are described and a new high-level parallel computing model called BSF is derived, which is an extension of the BSP model, and is based on the SPMD programming method and the «master-workers» framework. The BSF-model is oriented to computational systems with massive parallelism on distributed memory, including hundreds of thousands of processor nodes, and having an exaflop level of performance and numerical iterative methods with high time complexity. The BSF-computer architecture is defined, and the structure of the BSF-program is described. The formal cost metric, which provides parallel BSF-programs scalability upper bounds in context of distributed memory computing systems, is described. Also, formula for BSF-programs parallel efficiency are derived.