Experiences with Proxmox 1.9

F

Frank.Pawlik

Member
Oct 21, 2009
47
0
6
I have some negative experiences with Version 1.9.

1. Issues with pve-qemu-kvm (see my earlier post)

2. Free memory is not displayed correctly in the GUI.
I think it's because of anon hugespages in the new kernel.

3. Performance with RH-kernel (pve-kernel-2.6.32-6-pve) is subjective slower than with Debian-kernel (pve-kernel-2.6.32-4-pve).
Especially Windows 2008 R2 performs slower.
Maybe because of the now working cpuuinits or ksm.
I have to test.
 
Frank.Pawlik said:
2. Free memory is not displayed correctly in the GUI.
I think it's because of anon hugespages in the new kernel.
Click to expand...

What exactly is wrong? How does AnonHugePages influence the
value of MemFree in /proc/meminfo?


Frank.Pawlik said:
3. Performance with RH-kernel (pve-kernel-2.6.32-6-pve) is subjective slower than with Debian-kernel (pve-kernel-2.6.32-4-pve).
Especially Windows 2008 R2 performs slower.
Maybe because of the now working cpuuinits or ksm.
I have to test.
Click to expand...

Any numbers (benchmarks)?
 
At Point 2.

My error.
Seems to be correct.

Point 3:
At the Moment it is subjective.
Do you prefer some special benchmarks to compare with others?
At the host? At the vm?
 
can you run some phoronix benchmarks on Debian? you can run these benchmarks on the host, in a container, in a KVM guest. e.g. you can run the benchmark 'build-linux-kernel'

short howto for running phoronix on Debian:
List all available test:
Code: 
phoronix-test-suite list-tests

Execute a test, example:
Code: 
phoronix-test-suite benchmark build-linux-kernel

if you run 2.6.32-4 you will get always all CPU power in the container (as the cpu flag does not apply here). that means, if you got 16 cores on the host and you run a container with our defaults CPU=1 you will have 16 cores in the container - very fast, likely host performance. If you boot now 2.6.32-6, the cpu flag is used and you will got 1 CPU core in the container as specified - much slower, clearly. keep this in mind in your tests.
 
I have the first benchmarks (from our Testserver).

1. On the host with

phoronix-test-suite benchmark build-linux-kernel

Debian:
Timed Linux Kernel Compilation 2.6.25:
pts/build-linux-kernel-1.2.0
Test 1 of 1
Expected Trial Run Count: 3
Running Pre-Test Script @ 11:09:40
Started Run 1 @ 11:09:59
Running Interim-Test Script @ 11:34:05
Started Run 2 @ 11:34:11
Running Interim-Test Script @ 11:58:09
Started Run 3 @ 11:58:20 [Std. Dev: 0.17%]
Running Post-Test Script @ 12:22:19

Test Results:
1442.85702205
1438.42548609
1438.95309687

Average: 1440.08 Seconds

RedHat:
Timed Linux Kernel Compilation 2.6.25:
pts/build-linux-kernel-1.2.0
Test 1 of 1
Estimated Time Remaining: 1 Hour, 13 Minutes
Expected Trial Run Count: 3
Running Pre-Test Script @ 12:28:31
Started Run 1 @ 12:28:50
Running Interim-Test Script @ 12:50:11
Started Run 2 @ 12:50:19
Running Interim-Test Script @ 13:11:33
Started Run 3 @ 13:11:39 [Std. Dev: 0.10%]
Running Post-Test Script @ 13:32:55

Test Results:
1277.09111595
1274.508708
1275.76174593

Average: 1275.79 Seconds

2. In a XP - VM with SiSoft Sandra

No change, except HD is slightly faster with RH.


I will do some test inside Windows 2008 R2 as soon as possible.
The RH-Kernel seems to perform good.

Testserver:

Hardware:
Processor: Intel Xeon 3050 @ 2.13GHz (2 Cores), Motherboard: Intel S3000AHV, Chipset: Intel E7230/3000/3010 + ICH7, Memory: 4 x 2048 MB DDR2-533MHz, Disk: 500GB 9650SE-2LP DISK, Graphics: ATI ES1000, Network: Intel 82573E Gigabit

Software:
OS: Linux, Kernel: 2.6.32-6-pve (x86_64), Display Driver: ati, Compiler: GCC 4.3.2, File-System: ext3

pve-manager: 1.9-24 (pve-manager/1.9/6542)
running kernel: 2.6.32-6-pve
proxmox-ve-2.6.32: 1.9-46
pve-kernel-2.6.32-4-pve: 2.6.32-33
pve-kernel-2.6.32-6-pve: 2.6.32-46
qemu-server: 1.1-32
pve-firmware: 1.0-14
libpve-storage-perl: 1.0-19
vncterm: 0.9-2
vzctl: 3.0.29-1pve1
vzdump: 1.2-16
vzprocps: 2.0.11-2
vzquota: 3.0.11-1
pve-qemu-kvm: 0.15.0-1
ksm-control-daemon: 1.0-6

name: XPpro
ide2: none,media=cdrom
sockets: 1
vlan0: e1000=E6:85:FD:9C:14:83
ide0: local:101/vm-101-disk-1.qcow2
ostype: wxp
memory: 2048
onboot: 0
description:
boot: cad
freeze: 0
cpuunits: 10000
acpi: 1
kvm: 1
bootdisk: ide0
cores: 1
 
Last edited:
pls include details about your hardware setup.
 
I have the tests from W2k8 R2 now.

I tested with Hdbench, because Sandra is not running there.

Debian:
All
219255
CPU
Integer 599350 Float 616201
Memory
Read 406527 Write 420965 Read&Write 796569
HD
Read 242080 Write 198449 FileCopy 11347

RedHat:
All
195375
CPU
Integer 595074 Float 607878
Memory
Read 404078 Write 407810 Read&Write 779877
HD
Read 234324 Write 73668 FileCopy 11322

name: Prog
virtio0: LVM-VM-Disks:vm-108-disk-1,cache=none
ostype: w2k8
memory: 6144
sockets: 1
ide0: none,media=cdrom
virtio1: LVM-VM-Disks:vm-108-disk-2,cache=none
boot: c
freeze: 0
cpuunits: 25000
acpi: 1
kvm: 1
args: -cpu host
onboot: 1
description:
cores: 4
vlan1: virtio=56:3F:3A:2A:C4:6A
bootdisk: virtio0

pve-manager: 1.9-24 (pve-manager/1.9/6542)
running kernel: 2.6.32-6-pve
proxmox-ve-2.6.32: 1.9-43
pve-kernel-2.6.32-4-pve: 2.6.32-33
pve-kernel-2.6.32-6-pve: 2.6.32-43
qemu-server: 1.1-32
pve-firmware: 1.0-13
libpve-storage-perl: 1.0-19
vncterm: 0.9-2
vzctl: 3.0.28-1pve5
vzdump: 1.2-15
vzprocps: 2.0.11-2
vzquota: 3.0.11-1
pve-qemu-kvm: 0.15.0-1
ksm-control-daemon: 1.0-6

Hardware:
Intel Modular Server
MFS5520VI Compute Module
Intel(R) Xeon(R) CPU E5520 @ 2.27GHz 4 Cores
Seagate SAS ST9146803SS RAID 10
 

Attachments

  • hdbench_Debian.png
    hdbench_Debian.png
    12.9 KB · Views: 20
  • hdbench_RH.png
    hdbench_RH.png
    12.7 KB · Views: 13
Last edited:
It seems that RH performs better with Linux and slower with W2k R2.

Good news at the end.
Virtio-net is stable at last.
No need for e1000 any more.
 
You must log in or register to reply here.
Top Bottom