Load Balancing on Multimedia Client-Server Communication Networks: Computer Experiments

To design and implement an algorithm which, given the inputs of work cost, backlogs, and tasks for multiple servers, produces an output of work distributions (loads) for all servers and tasks in the system such that the time spans are minimal and, if possible, balanced. That is, the algorithm finds the optimal distribution for M tasks and N servers.
The project focuses on an algorithm for three server load balancing, and then attempts to generalize the algorithm to four and five servers. The system being considered consists of multiple servers represented as rows of a matrix, and multiple tasks, represented as columns of a matrix. Backlogs indicate the amount of work already being handled by a given server. Time spans indicate
the run time associated with running several tasks on a server. Tasks can be of any type of work; however, the algorithm is conceptually focused on multimedia tasks. The data initially has been provided as integers. The system is mathematically modeled as a system of linear inequalities, therefore it is a member of the “Linear Programming” class of problems.