Design and Implementation of vCPU Scheduling Architecture in Embedded Virtualization

This paper presents a novel scheduling architecture designed for the independently developed Reworks hypervisor (Revmm), aimed at enhancing scheduling efficiency through static precomputation. The architecture integrates a configuration generation mechanism, enabling scheduling algorithms to leverage precomputed static results to assist dynamic scheduling decisions, thereby minimizing runtime overhead. Experimental results demonstrate that the statically precomputed Round-Robin algorithm surpasses traditional Round-Robin scheduling in terms of runtime efficiency and stability. To meet real-time requirements, the study also introduces a static priority-based, preemptible multicore VCPU scheduling algorithm (SPPM) for Revmm. Experiments confirm that this algorithm ensures stable real-time performance for the highest-priority RTOS.