Testing XenServer filesystem performance on plain hardware and the difference between EXT and LVM.
How To Change the Default Storage Repository to EXT3
HardwareHardware
IBM Blade HS21 2x Intel(R) Xeon(R) CPU E5540 @ 2.53GHz 16 GB RAM LSI SAS 1064 with 2 SAS Disc with Raid 1
SoftwareSoftware
XenServer 5.6 SP2
Virtual Machines
Ubuntu VMs (64-Bit) with latest XenTools installed (1 CPU per VM and 1024 MB RAM)
Testsoftware
bonnie++
The Test
I run every test 3 times and used only the last two results. For example with 3 virtual machines i run bonnie++ 3 times on every virtual machine at the same time and stored the last 2 results of every virtual machine. (values tab in the table) I sum this 6 results and than divide it by 6. (summary tab in the table).
Download the OpenOffice table here:
Xen Performance Test (1307 Downloads )
Note: The data shows only the values for a single virtual maschine. For example the CPU Usage on the last tab in the table shows the CPU Usage of ONE virtual machine. (Multiplicate it with the amount of virtual machines and the difference between PLAIN is the virtualization overhead).
My result is, that there is no performance difference between LVM and EXT filesystem on the XenServer storage repository. But with EXT you can give the virtual maschines more storage as you physically have in your server.
Example:
gyrfalcon
17 März 2019Is this still valid? I was looking at different comparisons and trying to get more information on benefits or detractors from using LVM or EXT.
Also saw this article:
https://justyn.io/blog/xen-file-based-vs-lvm-based-disk-images-benchmarks/