An efficient feasibility test method for hard real-time periodic tasks

Addresses the problem of deciding the feasibility of hard real-time periodic tasks. It is known to be a co-NP problem to determine whether a task set is feasible on one processor when there exists a task with a relative deadline that is shorter than its period in the task set. For synchronous task sets, "processor demand analysis" (PDA) has been considered as a practical tool to solve the feasibility problem. PDA determines the feasibility of a task set by checking whether a deadline is missed in an interval of finite length; this time interval is called the "test interval". The efficiency of a feasibility test method depends on the length of the test interval. In this paper, we present a new method for the feasibility testing of hard real-time periodic tasks. We show theoretically that the length of the test interval in our algorithm is shorter than or equal to existing ones. We also present experimental results that show the length of the test interval in our algorithm is, on average, significantly shorter than existing ones.