HETEROGENEOUS MULTISPACE DATAFLOW NETWORK

Abstract

The article discusses a method for constructing a heterogeneous dataflow network based on the concept of a PF-network using the excess de Bruijn topology. The issues of the network structure, the main aspects of functioning, the principle of distribution of tasks and mechanisms for ensuring fault tolerance are considered. A superficial review of load balancing automation using tree decomposition and grain management was also done. A comparison with the original concept was made, the main gains, losses and prospects were identified.