The rapid growth of data-intensive applications has significantly increased the demand for memory capacity, driving the widespread adoption of the non-uniform memory access (NUMA) architecture. While NUMA offers significant advantages in memory capacity and performance, it presents challenges in resource scheduling, particularly in virtualized environments where the NUMA topology is often hidden from guest operating systems. We proposes a NUMA-aware cooperative resource scheduling scheme for multiple virtual machines (VMs) that addresses these challenges. Our approach replicates the host machine’s NUMA topology exactly in all VMs, enabling guest operating systems to fully exploit the underlying physical NUMA structure without requiring support for NUMA reconfiguration. We introduce three key components: A binding manager that exposes the NUMA topology to guest OSs, a balancing manager that dynamically allocates resources between VMs, and a custom memory allocator within the guest OS that enforces NUMA-aware memory allocation and migration. We are currently implementing these components and plan toevaluate their performance in the future.