Results from iozone testing on CentOS VM under Proxmox 2.0-45:
Proxmox kernel version: 2.6.32-10-pve; VM ( CentOS kernel info: 2.6.32)
Memory on VM is 6 GB, using virtio to a 100 GB LUN (RAID 0 over 4 disks) from the SAN for OS and disk space (together)
Note 1: Might want them to be separated for better performance
Note 2: Ran 4 runs of each test - get 2 results for reads/writes and then averaged all 8 numbers
Test line submitted is as follows: iozone -s #G -r 1024 -i 0 -i 1 -w -f <filename>
10 GB File (bigger than local RAM so did not use file sync option -o on test)
Write Ave= 48 MB/s
Read Ave= 183 MB/s
7 GB File (bigger than local RAM so did not use file sync option -o on test)
Write Ave= 41 MB/s
Read Ave= 297 MB/s
1 GB File (File test size smaller than local RAM, so USED file sync option -o on test)
Write Ave= 37 MB/s
Read Ave= 5.33 GB/s
1 GB File (Even though smaller than local RAM, DID NOT use file sync option -o on test)
Write Ave= 1.91 GB/s
Read Ave= 4.92 GB/s
Conclusion and Next Step:
I think that vast majority of file writes will be smaller than RAM so it would be valid to accept the Write Rates without file sync'ing as what would be observed in everyday usage.
Next step would be to run multiple VM's from Proxmox box and somehow come up with a loading test to see what kinds of rates can be realized.
Questions:
1- How do these results rate for this kind of configuration?
2- Is my conclusion valid for VM usage?
3- Does anyone have suggested other performance testing to evaluate how Proxmox will work under this configuration?
Proxmox kernel version: 2.6.32-10-pve; VM ( CentOS kernel info: 2.6.32)
Memory on VM is 6 GB, using virtio to a 100 GB LUN (RAID 0 over 4 disks) from the SAN for OS and disk space (together)
Note 1: Might want them to be separated for better performance
Note 2: Ran 4 runs of each test - get 2 results for reads/writes and then averaged all 8 numbers
Test line submitted is as follows: iozone -s #G -r 1024 -i 0 -i 1 -w -f <filename>
10 GB File (bigger than local RAM so did not use file sync option -o on test)
Write Ave= 48 MB/s
Read Ave= 183 MB/s
7 GB File (bigger than local RAM so did not use file sync option -o on test)
Write Ave= 41 MB/s
Read Ave= 297 MB/s
1 GB File (File test size smaller than local RAM, so USED file sync option -o on test)
Write Ave= 37 MB/s
Read Ave= 5.33 GB/s
1 GB File (Even though smaller than local RAM, DID NOT use file sync option -o on test)
Write Ave= 1.91 GB/s
Read Ave= 4.92 GB/s
Conclusion and Next Step:
I think that vast majority of file writes will be smaller than RAM so it would be valid to accept the Write Rates without file sync'ing as what would be observed in everyday usage.
Next step would be to run multiple VM's from Proxmox box and somehow come up with a loading test to see what kinds of rates can be realized.
Questions:
1- How do these results rate for this kind of configuration?
2- Is my conclusion valid for VM usage?
3- Does anyone have suggested other performance testing to evaluate how Proxmox will work under this configuration?