[Openstack] guest rbd block device speed limit

chagg at foxmail.com chagg at foxmail.com
Fri Aug 25 07:30:28 UTC 2017

I use this command to launch a VM. It disk io is also about 20M Bps,Why?

/usr/libexec/qemu-kvm -m 1024 -drive format=rbd,file=rbd:volumes/d9bda7d7-f68a-45ea-84a1-f50faa44b801_disk,cache=writeback -vnc :10 -usb -usbdevice tablet


chagg at foxmail.com
From: chagg at foxmail.com
Date: 2017-08-25 10:05
To: Chris Friesen; openstack
Subject: Re: Re: [Openstack] guest rbd block device speed limit
This is my ceph performance data:

 Total time run:         30.370630
Total writes made:      849
Write size:             4194304
Bandwidth (MB/sec):     111.819
Stddev Bandwidth:       21.6665
Max bandwidth (MB/sec): 124
Min bandwidth (MB/sec): 0
Average IOPS:           27
Average Latency(s):     0.572236
Stddev Latency(s):      0.421846
Max latency(s):         4.06547
Min latency(s):         0.121321


chagg at foxmail.com
From: Chris Friesen
Date: 2017-08-24 23:14
To: openstack at lists.openstack.org
Subject: Re: [Openstack] guest rbd block device speed limit
On 08/24/2017 01:04 AM, chagg at foxmail.com wrote:
> Hello:
>      I  am using openstack + libvirt + qemu-kvm. The speed of copying files
> between virtual machines exceeds 300M Byte per second but the speed of dd command:
>   "watch dd oflag=direct,nonblock if=/dev/zero of=/opt/iotest1 bs=4M count=10"
>   is around 20M byte per second. Every guest is the same, and there is no io
> tuning in libvirt. What can I do to unleash the speed of disk io speed inside
> the guest?
>      Thanks!
It looks like you're using ceph for your root disk.  What sort of performance do 
you get accessing the ceph volumes from the host?
Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to     : openstack at lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20170825/59d9f20e/attachment.html>

More information about the Openstack mailing list