Abstract: A method and apparatus for aggregating terminals into clusters that assists in the construction of a distributed data communication network. Terminal locations to be clustered are used as input along with a weight for each one representing the traffic carried by that location. The terminals are placed into boxes based on their geographic coordinates and only boxes with substantial traffic are retained. Any terminals included in a box not retained are assigned to the closest retained box. The retained boxes are merged into clusters so long as the maximum cluster traffic and maximum cluster radius are not exceeded. The center of mass of the resulting clusters are then determined. The representative locations of the resulting clusters are determined as the location of the terminal in the cluster closest to the center of mass.

Abstract: A method for constructing and utilizing a distributed data communication network that will meet its performance specifications in the most economical and efficient manner. The steps of the method include clustering terminals based on their proximity and traffic; selecting backbone node sites based on cost; designing the access network composed of terminals, concentrators, backbone nodes and the links between them; generating a backbone link topology; and determining routes in the backbone given the backbone topology and a traffic matrix.

Abstract: A method for selecting the most cost-efficient configuration for backbone links in a distributed data network is disclosed. The method identifies candidate links between pairs of backbone nodes, evaluates the effect on the network of adding each candidate link to the network, and adds to the network those candidate links which produce a cost savings in the network. The method also routes data through the backbone to minimize transmission delays, and eliminates redundant lines from underutilized links, and removes from the network those links whose removal reduces the cost of the backbone.

Abstract: A method to produce near-optimal routes for the flow of data between clusters in a distributed data communications network. A backbone traffic matrix and a backbone topology at the cluster level are used to produce minimum hop routes, minimum delay routes, or routes which maximize throughput. The inputs of the system are the number of requirements, the number of backbone links, and the number of backbone nodes.

Abstract: A method for selecting the most cost-efficient locations for backbone nodes in a distributed data network is disclosed. The method iteratively evaluates candidate backbone node locations, adding in each iteration the candidate whose addition to the network would create the greatest savings in total network cost. After all candidates which produce savings are added to the network, the method iteratively evaluates the backbone node locations and removes those whose removal would create cost savings. Data terminals are reassigned among the backbone nodes in the least cost manner. The method also estimates total network cost based on the resulting network configuration.

Abstract: A method for locating and selecting concentrators and local access links in a distributed data network in the most cost-efficient way is disclosed. The method first identifies a median concentrator from among a set of candidate concentrators to which all of the data terminals are initially assigned. The method then iteratively evaluates and adds to the network those concentrators which produce savings in the network's cost. The method first adds all of a set of concentrators specified to be mandatory and then adds from among a set of candidate concentrators those which produce cost savings. The method then evaluates concentrators in the network and removes those concentrators whose removal produces savings in the network cost.