Load adjustment and filtering based on process criticality

This paper generalizes the load scaling techniques proposed in (Kuo and Mok, 1991) for fixed-priority real-time applications. A framework is proposed to adjust the system workload by relating the criticality and flexibility of a process to the resource allocation problem. A load adjustment procedure based on the approximation algorithm (Ni et al., 1997) is proposed to maximize the system profit in an on-line fashion. When the list of allowable configurations is implicitly given by scalable periodic processes, the corresponding load filtering problem is shown to be NP-complete. We also prove the upper bound of the system profit and provide a mechanism to balance the schedulability and the maximum profit of the system.