Research on Resource Sharing and Scheduling Mechanism of Multi-Core Fusion System Based on Virtualization Technology

With the increasing demands for computing power in fields such as artificial intelligence and high-performance computing, heterogeneous multi-core fusion systems have gradually become the mainstream architecture. Addressing the inability of traditional scheduling mechanisms to meet the needs for high throughput and low latency, this paper proposes a multi-core resource sharing and scheduling mechanism based on virtualization technology. By implementing unified abstraction and collaborative scheduling of multi-core resources such as CPU, GPU, and FPGA on the KVM platform, this paper designs a shared page table and dynamic mapping mechanism to achieve efficient data sharing and consistency maintenance between virtual cores. At the same time, a heterogeneous-aware scheduling strategy is proposed, which dynamically adjusts task-resource matching to achieve efficient task migration and intelligent optimization of system response. Experimental verification shows that the proposed mechanism effectively improves resource utilization, reduces latency, and enhances the system's adaptability to burst loads in multi-core heterogeneous platforms. The results demonstrate that the mechanism has significant performance advantages and good scalability, providing both theoretical foundation and practical support for efficient scheduling and resource sharing in multi-core fusion systems.