Intelligent Scheduler for processing help requests on unrelated parallel machines in a computer support administration system

This paper describes an application-oriented examination of an automated online support process for the Center of Computing and Information Technology at The University of Arizona. The online consulting activity at the Center is modeled as a set of unrelated parallel machines (the consultants) which process a stream of incoming jobs (the online help request). Automatic indexing, a type of natural language processing, is applied to each job to estimate its machine dependent processing time. Following the time estimation scheme, we develop an online scheduling procedure whose objective is to minimize the mean flow time (reflecting perceived quality of service).