Federated Scheduling Optimization Scheme for Typed Tasks With Power Constraints in Heterogeneous Multicore Processor Architectures

In heterogeneous multicore processor architectures, it is a critical concern to optimize the performance of typed tasks (real-time and non-real-time tasks) under limited power consumption. In this paper, we propose a power-constrained federated scheduling optimization scheme for typed tasks based on both global and local scheduling systems. The global scheduling system adopts a federated scheduling strategy to split the typed task stream into multiple real-time and non-real-time sub-streams. The local scheduling systems are constructed as a set of M/M/c/m queueing systems with heterogeneous multicore processors considering task priority and finite queueing capacity, avoiding task blocking and resource wastage through finite cache and parallel execution of two types of task sub-flows. To meet the deadline constraints of real-time tasks, a real-time task queueing model with strong preemption priority is constructed, and a stable and efficient real-time scheduling algorithm based on sequential quadratic programming block homotopy is proposed, which can ensure the optimal distribution of real-time tasks while maintaining load balancing and schedulability. For non-real-time tasks, we propose a dichotomous search-based scheduling algorithm to minimize the average response time of non-real-time tasks under strong preemption constraints on real-time tasks, providing a theoretical analysis proof of the optimal processor speed configurations for heterogeneous multiprocessor systems under power constraints. Simulation results demonstrate that factors such as processor size, system capacity, available power, and task arrival rate have a significant impact on the system performance, and that the proposed scheme enables optimal load distribution for typed tasks with power constraints in multiprocessor queueing systems.