Server Virtualization: Para- and Full Virtualization: XenServer vs. KVM
El-Anani, Bashar Ramadan (2021)
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2021111120121
https://urn.fi/URN:NBN:fi:amk-2021111120121
Tiivistelmä
The purpose of this thesis was to implement testing environment through building bare metal Citrix XenServer and Kernel-based Virtual Machine (KVM) hypervisors and create one virtual machine on each hypervisor to examine the performance of full virtualization and paravirtualization technologies.
A seven-year-old tower server Fujitsu Primergy TX150 S8 with Xeon E5-2407 2.2 GHz processor, 16 GB RAM, 145 GB hard drive and one 1 Gb NIC was used to host hypervisors. Different virtual machine managers were introduced and discussed. After virtual machines creation, Phoronix test suite program was used to test processors and memory performance, while iPerf software was used for network performance tests.
Performance test results were collected and analyzed. Citrix XenServer virtual machine results showed better performance scores than KVM results.
In conclusion, the results proved that full virtualization performance is slower than paravirtualization performance. Although the bare metal hypervisors were built successfully and reflected good performance, it is not recommended to seek only free, used and old technology solutions, because much time would be wasted on deploying and maintaining the system. This project is suitable for IT learners and small companies’ internal operations.
A seven-year-old tower server Fujitsu Primergy TX150 S8 with Xeon E5-2407 2.2 GHz processor, 16 GB RAM, 145 GB hard drive and one 1 Gb NIC was used to host hypervisors. Different virtual machine managers were introduced and discussed. After virtual machines creation, Phoronix test suite program was used to test processors and memory performance, while iPerf software was used for network performance tests.
Performance test results were collected and analyzed. Citrix XenServer virtual machine results showed better performance scores than KVM results.
In conclusion, the results proved that full virtualization performance is slower than paravirtualization performance. Although the bare metal hypervisors were built successfully and reflected good performance, it is not recommended to seek only free, used and old technology solutions, because much time would be wasted on deploying and maintaining the system. This project is suitable for IT learners and small companies’ internal operations.