[Gluster-users] Write performance on XenServer guest

[Stefano Baronio]

Here are the tests:

Gluster 3.1.2 servers are CentOS with 4Gb RAM and RAID1 (system) +
RAID5 3HD 74Gb
Tests ran from XenServer Dom0 (CentOS like system)

NFS standard
Write:
time sh -c "dd if=/dev/zero
of=/var/run/sr-mount/db70c449-9ff5-e28c-73ec-25f4fc0b07cf/test/file1.raw
bs=512KB count=9765 && sync"
9765+0 records in
9765+0 records out
4999680000 bytes (5.0 GB) copied, 63.675 seconds, 78.5 MB/s
real 1m4.118s
user 0m0.000s
sys 0m7.700s

Read:
time sh -c "dd if=/var/run/sr-mount/db70c449-9ff5-e28c-73ec-25f4fc0b07cf/test/file1.raw
of=/dev/null bs=512KB count=9765 && sync"
9765+0 records in
9765+0 records out
4999680000 bytes (5.0 GB) copied, 68.3861 seconds, 73.1 MB/s
real	1m8.939s
user	0m0.040s
sys	0m4.540s


gluster 3.1.2 replicate mode

Write
time sh -c "dd if=/dev/zero
of=/var/run/sr-mount/db70c449-9ff5-e28c-73ec-25f4fc0b07cf/test/file1.raw
bs=512KB count=9765 && sync"

9765+0 records in
9765+0 records out
4999680000 bytes (5.0 GB) copied, 152.735 seconds, 32.7 MB/s

real 2m33.049s
user 0m0.000s
sys 0m9.950s


Read
time sh -c "dd if=/var/run/sr-mount/db70c449-9ff5-e28c-73ec-25f4fc0b07cf/test/file1.raw
of=/dev/null bs=512KB count=9765 && sync"

9765+0 records in
9765+0 records out
4999680000 bytes (5.0 GB) copied, 61.8462 seconds, 80.8 MB/s

real 1m2.233s
user 0m0.000s
sys 0m4.300s


Thank you
Stefano



2011/1/27 Shehjar Tikoo <shehjart at gluster.com>
>
> Shehjar Tikoo wrote:
>>
>> What gluster config are you using?
>
>
> My bad.. I didnt read the mail completely. First thing you should try is run a streaming IO write perf test using dd or iozone. Lets see how that performs over the replicated config. Thanks
>
>
>
>>
>
>> Stefano Baronio wrote:
>>>
>>> Hello,
>>>    I have recently set up gluster-3.1.2 as NFS Virtual Disk Storage for
>>> XenServer.
>>> I have ran a windows vm on it and tested the disk performance:
>>> Read: 100MB/s
>>> Write: 10MB/s
>>>
>>> While with standard NFS, on the same servers, we can achieve:
>>> Read: 115MB/s
>>> Write: 100MB/s
>>>
>>> We have two servers with local scsi disks with gluster configured in
>>> replicate mode.
>>> Volume Options:
>>>
>>> performance.cache-size 1GB
>>>
>>> write-behind-window-size 512MB
>>>
>>> performance.stat-prefech 1
>>>
>>> Is there any specific option for this configuration? And, is it normal such
>>> a poor write performance?
>>>
>>>
>>> Thank you
>>> Stefano Baronio
>>>
>>>
>>>
>>> ------------------------------------------------------------------------
>>>
>>> _______________________________________________
>>> Gluster-users mailing list
>>> Gluster-users at gluster.org
>>> http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
>>
>> _______________________________________________
>> Gluster-users mailing list
>> Gluster-users at gluster.org
>> http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
>