NUMA-Aware PER-CPU Framework: Enhancing Latency and Storage Efficiency