Authors: REFİK SAMET, ORHAN FİKRET DUMAN
Abstract: In this article we present an analysis for task models having random resource needs and different arrival patterns. In hard real-time environments like avionic systems or nuclear reactors, the inputs to the system are obtained from real world by using sensors. And it is highly possible for a task to have different resource needs for each period according to these changing conditions of real world. We made an analysis of schedulers for task models having random resources in each period. Since feasibility tests for usual task models are just limited to some specific schedulers and arrival patterns, we made our analysis using different preemptive schedulers with different arrival patterns. Another issue that this paper presents is that for task sets having processor utilization factors of below 100%, a new scheduler assigns more resource values to the tasks than they originally have by extending and shrinking the resource values. This provides a better utilization of the resources as well as providing all the tasks to finish in their timing constraints and never assigning lower resource values than the initial resources for all tasks. We exposed a dynamic priority driven scheduling scheme that implements the newly developed Steady Scheme with Bounded Utilization for changing resources and simulated to reach the results depicted.
Keywords: Scheduling, real-time scheduler, task arrival patterns, simulation, dynamic priority, random resource, resource modification
Full Text: PDF