| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 2.07 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
With the hardware breakthroughs accomplished through the years, the idea of software defined hardware has become a reality. Hypervisors such as KVM, Xen, Hyper-V and ESXi enable the cloud of today, with hardware consolidation bringing a reduction in operating costs. In this scope, it is imperative to address the performance of all the different virtualization implementations, in order to discover any potential bottlenecks and bugs. In this work, the performance of all the prominent Type-1 virtualization platforms is analyzed, using guests representative of the Windows NT and Linux kernels, in the form of Windows 10 LTSB and Ubuntu Server 16.04 LTS. The effectiveness of the CPU scheduler of each hypervisor is put to the test, as well as the storage backend performance under multiple scenarios (iSCSI, NFS and local). In short, this project provides a snapshot of the current state of the virtualization market, covering CPU, Memory, 2D & 3D Graphics performance of oVirt, Proxmox, XenServer, Hyper-V and VMware Vsphere. All the benchmarks were executed using their own default settings, with some automation scripts, in order to accelerate the process and exclude variability as much
as possible. Among the selected benchmarks were: Passmark Performance Test 9 to benchmark Windows performance; Unixbench, providing a way to extrapolate the performance of Linux guests; (ez)FIO allowed in-depth analysis of filesystem performance across platforms. Concluding, there are a few generalizations that can
be made from the information gathered: XenServer, oVirt and Proxmox require the presence
of xentools/virtio in order to provide good I/O throughput; GPU passthrough provides native performance as
long as there is no resource overcommitment; VMware's Vsphere provides impressive CPU performance, edging out the competition, with 98% of the native performance; Hyper-V offers mediocre 2D Desktop performance (28% of the native performance), as such, it should not be used in VMs that provide interactive desktops; Similarly, Hyper-V's performance plunges in memory related workloads, when compared to the
remaining platforms and bare metal, with a mere 83%; The remote I/O results crown iSCSI as best performer, with double the performance of NFS; All the open source platforms (Proxmox, oVirt and XenServer) display
impressive remote I/O performance, in both iSCSI and NFS.
Description
Keywords
Bare metal virtualization Benchmarking Performance assessment
Citation
Pousa, Duarte; Rufino, José (2017). Practical study of bare metal virtualization solutions. In eji v Encontro de Jovens Investigadores. Instituto Politécnico de Bragança
Publisher
Instituto Politécnico de Bragança