CPU and memory performance analysis on dynamic and dedicated resource allocation using XenServer in Data Center environment

Abstract

Idle servers only utilize 10% to 20% of their potential. As such, configuring computer resource management has been considered to optimize servers' capabilities. In this research, a series of experiments were conducted using virtual machines (VMs) with different CPU and memory allocation techniques implemented in XenServer. The created virtual machines ran benchmarks and stress test applications with a preconfigured test and duration to understand the virtual machines' performance when given a continuous load. These applications consist of CPU, memory, disk intensive app as well as benchmark apps. The performance parameters in this research were the benchmark apps results. These experiments indicate that the best CPU allocation for file transfer and MySQL query tests are Priority Weight and Cap CPU allocation, with performance differences of around 75% compared to Default CPU allocation. Meanwhile, the CPU allocation method for Stress-ng CPU stress test, Stress-ng VM stress test, and UnixBench tests is the Default CPU allocation with a performance difference of around 45%. Furthermore, Default CPU allocation is the best method of CPU allocation for web servers with performance differences of between 5% to 29% over Priority Weight and Cap CPU allocation. The best memory allocation method is Static Memory allocation, with a performance difference between 7% to 53% over Dynamic Memory allocation in all tests.