Abstract:
Effective aggregation of separate (super)computers requires the use of dynamic parallelization technology, automatic dynamic reconfiguration of computing field, dynamic load balancing, as well as support fault tolerance, taking into account the reliability of nodes and the network topology. The article discusses the domestic software technology developed under the above requirements in the program SKIF-GRID.
Key words and phrases:cloud computing, distributed computing, GRID systems, dynamic program paralleling, T-system with an open architecture, fault-tolerance.