Optimizing Huge Page Allocation for Efficient Memory Management in Virtualized Systems

Optimizing memory utilization is crucial for achieving efficient performance in virtualized systems. This paper explores the allocation of Huge Pages, larger page sizes, to improve memory management within virtual machines (VMs). The experiment focuses on determining the optimal number of Huge Pages for VMs based on their memory requirements. The paper presents a formula for calculating the number of Huge Pages to be allocated, taking into account the total memory assigned to VMs and a buffer value. By applying this formula, the experiment successfully demonstrates the reduction of unused Huge Pages and optimized memory usage in the host system. Additionally, the paper benchmarks the performance of system memory within the VM environment using the Sysbench benchmarking tool. The experiment compares the latency results of lightweight and high memory usage tests across different Huge Pages allocations. The findings highlight the negligible impact of Huge Pages size on latency in the lightweight memory test, while emphasizing the importance of appropriate Huge Pages allocation for memory-intensive workloads. The results provide valuable insights into memory performance and optimization in virtualized environments. The findings contribute to improving overall system efficiency and resource utilization. Future plans involve further exploration of memory-intensive workloads and investigation of alternative memory management techniques, such as Transparent Huge Pages (THP).