[one-users] A question about the utilization of VMs

[Adnan Pasic]

Okay, I found out additional infos regarding my problem. As it seems, the
fact that the libvirt’s “ksm” dynamically manages the memory pages, which
obviously leads to a stabilization of used memory, when the VM’s virtual
memory is utilized to the maximum. This is why I deactivated “ksm” and
voila: Now the real memory stays utilized as well as the virtual memory, and
doesn’t handle “clean” memory pages anymore.

However, I have another problem now – after I finish the utilization of my
VM (which means, I kill the util. process, thus making all my virtual memory
free again), the real host memory still stays utilized.

What could be the reason that my VM shows 0% mem. util., while the physical
host still has 100%??? 

I thought maybe the reason could be that ONE accesses the “virto_balloon”?
Can the authors maybe give me an answer to this?

All in all, is there any way to make this work – to fully utilize a VM for
10 min. and the host stays utilized as well, and when I stop the utilization
that both the VM and host memory decrease???

 

Thanks!

 
 
 
________________________________
 Von: Steven Timm <
<http://lists.opennebula.org/listinfo.cgi/users-opennebula.org> timm at
fnal.gov>
An: Adnan Pasic <
<http://lists.opennebula.org/listinfo.cgi/users-opennebula.org> pqado at
yahoo.de> 
Cc: " <http://lists.opennebula.org/listinfo.cgi/users-opennebula.org> users
at lists.opennebula.org" <
<http://lists.opennebula.org/listinfo.cgi/users-opennebula.org> users at
lists.opennebula.org> 
Gesendet: 16:08 Mittwoch, 23.November 2011
Betreff: Re: [one-users] A question about the utilization of VMs
 
What hypervisor are you using? If you are using KVM then this
behavior is normal, KVM only actually occupies the amount of RAM
that the virtual machine is actually using at the time and
OpenNebula reports the usage accordingly.  this makes it possible
to overbook and put 16 2-GB KVM VM's on a machine that has less physical
RAM than that.
 
Steve Timm
 
 
On Wed, 23 Nov 2011, Adnan Pasic wrote:
 
> Hello guys,
I have a
problem with understand one specific behaviour and I hope that this
community
might help me with it.
Anyway, I
am writing my diploma thesis using OpenNebula and so far it has done
everything
I imagined and wanted it to do. I deploy VMs and utilize every one of them
with
the tool ÿÿlookbusyÿÿ. On normal physical hosts, lookbusy works perfectly,
utilizing the CPU/Memory as I set it and doesnÿÿt stop the utilization until
the
whole process is killed.
When I
utilize my VM, however, I noticed the VMs being utilized for approx. 2
minutes,
but afterwards I can see a falling trend of utilized memory on the physical
host. For instance, I utilize the memory of the VM by 2 GB ÿÿ thus, the
physical
host shows the same utilization too. Nevertheless, after 2 minutes, the
utilization on the physical host starts decreasing and stops at about 500MB
of
utilized memory, although the VM shows the whole time a full utilization of
2GB
!
Do you know
what could be the reason? Maybe it has something to do with the scheduler,
or I
donÿÿt know?
Please help
me with this, as this is a crucial part of my diploma thesis and I got
stuck!!!
 
Thanks in
advance.
Regards,
Adnan

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opennebula.org/pipermail/users-opennebula.org/attachments/20111123/66761b47/attachment-0003.htm>