Experience of multilevel parallelizing of the branch and bound method in discrete optimization problems

Various schemes are considered of the parallel implementation of the branch and bound method, as applied to multiprocessor computing systems (clusters) with the distributed memory. In the language of informal automata, questions are set out of the organization of the exchange of data and signals within the cluster, which afford the asynchronous operation of its processors. Common ideas are illustrated by the example of the classical traveling salesman problem and data of numerical experiments performed on the multiprocessor computing system-100 (MCS-100) are given.