Worst-case analysis for on-line service policies

This paper considers an on-line scheduling problem occurred in a service system with two classes (ordinary and special) of customers and two types (dedicated and flexible) of servers. The customers line up with a list according to the order of their arrivals. All customers are assumed to be in the system when service starts and no preemption is allowed. The objective is to complete all services in list as early as possible. We study three classes of service policies that are commonly used in practice and analyze their worst-case performance in terms of makespan. Tight worst-case ratios have been derived for these service policies.