(Mona Abou-Sayed, Charlene Edwards, Scott Singleton, 1995)
The thrust of this report is to present pertinent, detailed information regarding the network design for a client of MCI Communications Corp. Within this report we plan to present our findings, which include the optimum number of nodes and where they should be placed, as well as how and why this number of nodes and placement of nodes was obtained. In addition, we will show how many DS-3 lines need to be placed to connect all the cities with demand.
In finding a solution for the Digital Signal 3 network, many avenues were ventured, some of which failed and others which succeeded. One approach involved a program by Professor Barr (SMU) that showed the demand for a particular circuit path and compared any circuit path to another, giving the conflicting times between the two paths. Also used was a heuristic measurement we called ``radius of access". This measurement helped us find the initial placement of nodes and started us with analyzing cost minimization. With the purchase of a U.S. map we were able to pin point the first set of initial nodes which included: Atlanta, Washington, New York and Los Angeles. Several discoveries were made in utilizing the map and thus started us to re-evaluate our initial node assumptions. After this evaluation our new set of nodes included: Atlanta, Albuquerque, Fresno, Omaha and Philadelphia.
Our next attempt was to create a Cplex (optimization) program to model our network. In order to do this we had to first try to solve for the easiest case scenario which involved circuits that contained cities reaching only one node. In solving this scenario, we were able to determine the number of DS-3 lines needed to connect nodes to cities. Unable to obtain a working Cplex program, a working model was done by hand. From this hand model we were able to calculate all node to destination costs and determine the number of DS-3 lines needed for the failure contingency.
We also consolidated a node to optimize costs, which again caused us to change the node set to the final recommendation: Dallas, Los Angeles, Tampa, Atlanta, Omaha and Philadelphia. Consolidation of nodes and demand involves placing nodes in geographically advantageous locations, which helps to cut cost over time. The optimal cost was found to be: $4,461,868.50. This figure could stand to be improved with further consolidation; however, due to time constraints we were unable to continue with minimizing of costs.
Upon completion of this project, we were able to come up with a generic Cplex model for the entire network. A small sample of only six demand circuits was tested on Cplex. This sample was found generated the desired output, which could be expanded to solve the network problem with additional attention.