ИСТИНА

Computational network (CN) is a set of computer nodes connected by a data transfer network and cooperatively processing a stream of incoming tasks (workload). Timely execution of tasks and balanced load of nodes are both required for CN operation. In this paper, we propose a machine learning-based multi-agent algorithm for task allocation to CN nodes. The convergence and solution quality of the algorithm are experimentally studied for different CN topologies and workloads. The algorithm requires splitting the CN into several domains with a dedicated agent in every domain. Influence of the number of domains on algorithm convergence is also experimentally studied.