Self-adaptive teleprocessing network design

Abstract

A technique is presented for designing teleprocessing networks on the basis of weighted heuristic partitions of the network nodes. The partitioning groups the nodes on the basis of common internodal design characteristics related to traffic flow, geographical location, delay and reliability requirements. The research results include a theoretical framework for performing the heuristic partition selection and for analyzing the resulting partition set to produce a complete network design. Both techniques are based on the theory of connectivity in directed graphs. The techniques are implemented in the form of a complete network design system using a set of APL functions. Weights incorporated into the partition selection functions permit adaptive adjustment of the design process based on actual experience gained in designing networks. Several complete examples are included, as is a sensitivity analysis of the partition selection functions and weights.