cloudstack-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Carlos Reátegui <create...@gmail.com>
Subject Re: Disk IO on Cloudstack VMs
Date Fri, 07 Mar 2014 04:32:16 GMT
Nishan,
How did you originally setup your networks?  Did you do it all through XenCenter, xe CLI,
OS?  I believe I did it all through XC.  I am running XS 6.0.2 with CS 4.1.  Did you switch
to bridge networking?  I did not as it only supports 2 nic bonds.

My hosts are set up with 4x1Gbe bond for NAS network (storage label) and 4x1Gbe bond for CS
Management/CS Storage/CS Guest (cloud-public label).  Both bonds are active active.  I am
using basic networking but without security groups since that requires bridge networking.
 Security is not really an issue for this setup as it is used internally by QA and dev.

On one of my hosts:
# dd if=/dev/zero of=test.dmp bs=1M conv=fdatasync count=1000
1000+0 records in
1000+0 records out
1048576000 bytes (1.0 GB) copied, 9.97504 seconds, 105 MB/s

On a CentOS 5 guest:
# dd if=/dev/zero of=/mnt/test.dmp bs=1M conv=fdatasync count=1000
1000+0 records in
1000+0 records out
1048576000 bytes (1.0 GB) copied, 13.1997 seconds, 79.4 MB/s

On an ubuntu 12.04 guest:
$ dd if=/dev/zero of=test.dmp bs=1M conv=fdatasync count=1000
1000+0 records in
1000+0 records out
1048576000 bytes (1.0 GB) copied, 9.85764 s, 106 MB/s

I tried the above a handful of times on each and they were within a couple % of each other
every time.  Both of the guests are on the same host I tested above.  The SR is on NFS on
the same machine as the CS Management stack (Ubuntu 12.04).  That machine is connected to
the storage via 10Gbe and a 2x1Gbe bond to the cloud-public network.  No routes exist between
these 2 networks.  MTU is 1500 for both networks.  I have another setup at a different location
where I set the MTU to 9000 and I seem to remember getting a bit faster throughput at least
on the hosts.  I’ll check later and let you know.

Let me know if there is anything you want me to check in my setup to compare.

Carlos


On Mar 6, 2014, at 5:14 PM, Nishan Sanjeewa Gunasekara <nishan.sanjeewa@gmail.com> wrote:

> Hi Shankar,
> 
> 10% overhead for virtual disk seems reasonable.
> My VM is PV and I've installed xenserver tools as well. But still getting
> 20Mbps where as on the host its 150Mbps.
> 
> Have you tuned up your VM or changed any parameters?
> 
> Regards,
> Nishan
> 
> 
> On Wed, Mar 5, 2014 at 4:15 PM, Shanker Balan
> <shanker.balan@shapeblue.com>wrote:
> 
>> Comments inline.
>> 
>> On 05-Mar-2014, at 10:01 am, Nishan Sanjeewa Gunasekara <
>> nishan.sanjeewa@gmail.com> wrote:
>> 
>>> Hi,
>>> 
>>> I'm running CS 4.2.1 on XenServer 6.2 hosts.
>>> 
>>> My primary storage is a ZFS file system provided via NFS.
>>> 
>>> When I do a dd test directly from one of the hosts on the NFS mount I
>> get a
>>> write speed of about 150MBps
>>> 
>>> dd if=/dev/zero of=test.dmp bs=1M conv=fdatasync count=1000
>>> 1000+0 records in
>>> 1000+0 records out
>>> 1048576000 bytes (1.0 GB) copied, 6.99829 seconds, 150 MB/s
>>> 
>>> But when I do the same test on a Cloudstack VM running on the same host
>>> (root disk on the same nfs mount ofcourse) I get a very low write speed.
>>> 20MBps.
>>> 
>>> dd if=/dev/zero of=test.dmp bs=1M conv=fdatasync count=1000
>>> 1000+0 records in
>>> 1000+0 records out
>>> 1048576000 bytes (1.0 GB) copied, 44.6171 s, 23.5 MB/s
>>> 
>>> Any ideas how I can improve this ?
>> 
>> 
>> Is this a PV VM?
>> 
>> I just did a unscientific test in my lab on a PV VM. Results below:
>> 
>> On the XenServer:
>> 
>> [root@vxen1-1 ebb66062-d46f-7b3a-07be-b9ec583ec1a9]# dd if=/dev/zero
>> of=test.dmp bs=1M conv=fdatasync count=1000
>> 1000+0 records in
>> 1000+0 records out
>> 1048576000 bytes (1.0 GB) copied, 26.1396 seconds, 40.1 MB/s
>> 
>> On a VM:
>> 
>> [root@scan1 ~]# dd if=/dev/zero of=test.dmp bs=1M conv=fdatasync
>> count=1000
>> 1000+0 records in
>> 1000+0 records out
>> 1048576000 bytes (1.0 GB) copied, 29.4688 s, 35.6 MB/s
>> 
>> 
>> I guess 10% is the virtual disk overhead.
>> 
>> 
>> 
>> 
>> --
>> @shankerbalan
>> 
>> M: +91 98860 60539 | O: +91 (80) 67935867
>> shanker.balan@shapeblue.com | www.shapeblue.com | Twitter:@shapeblue
>> ShapeBlue Services India LLP, 22nd floor, Unit 2201A, World Trade Centre,
>> Bangalore - 560 055
>> 
>> Need Enterprise Grade Support for Apache CloudStack?
>> Our CloudStack Infrastructure Support<
>> http://shapeblue.com/cloudstack-infrastructure-support/> offers the best
>> 24/7 SLA for CloudStack Environments.
>> 
>> Apache CloudStack Bootcamp training courses
>> 
>> **NEW!** CloudStack 4.2.1 training<
>> http://shapeblue.com/cloudstack-training/>
>> 18th-19th February 2014, Brazil. Classroom<
>> http://shapeblue.com/cloudstack-training/>
>> 17th-23rd March 2014, Region A. Instructor led, On-line<
>> http://shapeblue.com/cloudstack-training/>
>> 24th-28th March 2014, Region B. Instructor led, On-line<
>> http://shapeblue.com/cloudstack-training/>
>> 16th-20th June 2014, Region A. Instructor led, On-line<
>> http://shapeblue.com/cloudstack-training/>
>> 23rd-27th June 2014, Region B. Instructor led, On-line<
>> http://shapeblue.com/cloudstack-training/>
>> 
>> This email and any attachments to it may be confidential and are intended
>> solely for the use of the individual to whom it is addressed. Any views or
>> opinions expressed are solely those of the author and do not necessarily
>> represent those of Shape Blue Ltd or related companies. If you are not the
>> intended recipient of this email, you must neither take any action based
>> upon its contents, nor copy or show it to anyone. Please contact the sender
>> if you believe you have received this email in error. Shape Blue Ltd is a
>> company incorporated in England & Wales. ShapeBlue Services India LLP is a
>> company incorporated in India and is operated under license from Shape Blue
>> Ltd. Shape Blue Brasil Consultoria Ltda is a company incorporated in Brasil
>> and is operated under license from Shape Blue Ltd. ShapeBlue is a
>> registered trademark.
>> 


Mime
View raw message