M. Tech. (Department of Information Communication Technology,
Faculty of Applied and Computer Sciences), Vaal University of Technology. / Virtualization is the main technology that powers today’s cloud computing systems. Virtualization provides isolation as well as resource control that enable multiple workloads to run efficiently on a single shared machine and thus allows servers that traditionally require multiple physical machines to be consolidated to a single, cost-effective physical machine using virtual machines or containers. Due to virtual machine techniques, the strategies that improve performance like hardware acceleration, running concurrent virtual machines without the correct proper resource controls not used and correctly configured, the problems of scalability as well as service provisioning (crashing response time, resource contention and functionality or usability) for cloud computing, emanate from the configurations of the virtualized system. Virtualization performance is a critical factor in datacentre and cloud computing service delivery. To evaluate virtualization performance as well as to determine which virtual machine configuration provides effective performance, how to allocate and distribute resources for virtual machine performance equally is critical in this research study. In this study, datacentre purposed servers together with Type 1 (bare metal hypervisors), VMware ESXi 5.5, and Proxmox 5.3 were used to evaluate virtualization performance. The experimental environment was conducted on server Cisco UCS B200 M4 which was the host machine and the virtual environment that is encapsulated within the physical layer which hosts the guest virtual machines consisting of virtual hardware, Guest OSs, and third-party applications. The host server consists of virtual machines with one operating system, CentOS 7 64 bit. For performance evaluation purposes, each guest operating system was configured and allocated the same amount of virtual system resources. Various Workload/benchmarking tools were used for Network, CPU, Memory as well as Disk performance, namely; Iperf, Unibench, Ramspeed, and IOzone, respectively. In the case of Iozone, VMware was more than twice as fast as Proxmox. Although CPU utilization in Proxmox was not noticeably affected, considerably less CPU utilization was observed in VMware. While testing the memory performance with ramspeed, VMware performs 16 to 26% better than Proxmox. In the case of writing, VMware observed 31 to 51% better than Proxmox. In a network, it was observed that the performance on Proxmox was very close to the level of bare metal setup. The results of the performance tests show that the additional operations required by virtualization can be confirmed utilizing test programs. The number of additional operations and their type influence specifically to performance as overhead. In memory and disk areas, where the virtualization
procedure was clear, the test outcomes demonstrate that the measure of overhead is little. Processor and network virtualization, then again, was more perplexing. Hence the overhead is more significant. At the point when the overall performance of a virtual machine running in VMware ESXi Server is contrasted with a conventional system, the virtualization causes approximately an increase of 33% in performance.Because of the difficulty in providing optimal real system configurations, workload/benchmarks could provide close to real application systems for better results. The tests demonstrate that virtualization depends immensely on the host system and the virtualization software. Given the tests, both VMware ESXi Server and Proxmox are capable of providing Optimal performance.
Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:vut/oai:digiresearch.vut.ac.za:10352/489 |
Date | 04 October 2019 |
Creators | Thovheyi, Khathutshelo Nicholas |
Contributors | Zuva, Tranos, Prof., Okosun, Kazeem, Prof. |
Publisher | Vaal University of Technology |
Source Sets | South African National ETD Portal |
Language | English |
Detected Language | English |
Type | Thesis |
Page generated in 0.0017 seconds