VCSRT extended page table assisted virtual memory mapping framework for efficient dynamic resource allocation in industrial edge virtualized runtimes

Abstract

The shift toward the Industrial Internet imposes new requirements on industrial edge runtime systems for rapid resource configuration and dynamic application deployment. However, existing embedded and real-time runtimes suffer from low efficiency and redundant logic in dynamic physical memory management, which can hardly meet the deterministic demands of industrial mixed-criticality scenarios. To address these limitations, this paper proposes VCSRT, a hardware-assisted virtual memory mapping framework tailored for industrial edge scenarios. It integrates three core components: an optimized Lazy Buddy system for on-demand memory allocation, Extended Page Table (EPT)-based hardware-accelerated address translation, and priority-aware scheduling adapted to industrial workloads. By eliminating the architectural redundancy of conventional virtual memory stacks, VCSRT substantially improves memory allocation efficiency and deterministic resource provisioning. Experimental results validate that VCSRT achieves substantial reductions in worst-case allocation latency and worst-case execution time of high-priority tasks under heavy load, and maintains microsecond-level response jitter stability across all load levels. The proposed framework can provide efficient and deterministic memory resource support for industrial edge dynamic reconfiguration applications.