Maintaining temporal validity of real-time data in component-based systems

How to schedule the sensor transactions to maintain the temporal validity of real-time data is an important research issue for cyber-physical systems. Current studies on sensor transaction scheduling assume that the sensor transactions are executed on a platform with continuous resource supply. This assumption does not hold for component-based systems where each component's transactions obtain a fraction of the resources that are usually supplied by the underlying platform in a non-continuous manner. In this paper, we study the problem of scheduling sensor transactions in component-based systems. The sensor transactions are assumed to be executed in one component of the system that is deployed on a multiprocessor platform. The resource supply of each processor follows the explicit-deadline periodic (EDP) resource model. A partitioned scheduling method named PS-FC is proposed at first. It uses a sufficient and necessary condition of EDF scheduling upon an EDP resource to check the feasibility of each transaction on a processor. Then two faster partitioned scheduling methods, named PS-CI and PS-Hybrid, are proposed. PS-CI does the feasibility checks for transactions based on the candidate intervals of transaction deadlines. PS-Hybrid is a combination of PS-FC and PS-CI. The effectiveness and efficiency of the methods are evaluated through extensive experiments.