> One more question actually, this machine is gonna be a NFS "homes"
> server for 30 Linux workstations. The homes are fairly IO busy,
> especially when rsnapshot runs :) The question is, would such a
> workload be happy to run in a Xen domU or should I be running it as
> the dom0 ? I think it makes more sense to use it as the dom0!

You should run the NFS server in a DomU. The Dom0 isn't "special" -- don't confuse it with the Xen hypervisor, which runs *under* all of the domains, even Dom0. Dom0, for security purposes, should be locked down extremely tight. Any services that your box provides should be served by domU's, not Dom0. This is because anybody that can gain access to your Dom0 can manipulate your DomU's, which is a huge security problem.

Paul Krizak                         7171 Southwest Pkwy MS B400.2A
Advanced Micro Devices              Austin, TX  78735
Linux/Unix Systems Engineering      Desk:  (512) 602-8775
Silicon Design Division             Cell:  (512) 791-0686


Ahmed Kamal wrote:
Cool, thanks a lot. Would have been cool for xm to offer "swap" like
features for over committing RAM. Anyway, thanks a million

One more question actually, this machine is gonna be a NFS "homes"
server for 30 Linux workstations. The homes are fairly IO busy,
especially when rsnapshot runs :) The question is, would such a
workload be happy to run in a Xen domU or should I be running it as
the dom0 ? I think it makes more sense to use it as the dom0!

Thanks and Best Regards


On Mon, Jul 21, 2008 at 11:48 AM, Zavodsky, Daniel (GE Money)
<[EMAIL PROTECTED]> wrote:
Hello,
       You cannot overcommit memory. However, you can set max-mem to a
higher value and then dynamically (but manually) control the memory
allocation to individual domains.
       You can also control the amount of CPU power a domain is
guaranteed to get via the xm sched-credit command. You assign weights to
domains, which are relative.
       For example if Dom1 and Dom2 should get both 30 percent and Dom0
should get 40, you would set it for example thus:

xm sched-credit -d Dom1 -w 300
xm sched-credit -d Dom2 -w 300
xm sched-credit -d Dom0 -w 400


In other cases, domains can eat up to the assigned number of VCPUs of
CPU power.


Regards,
       Daniel


-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Ahmed Kamal
Sent: Monday, July 21, 2008 10:38 AM
To: Red Hat Enterprise Linux 5 (Tikanga) discussion mailing-list
Subject: [rhelv5-list] guaranteed cpu power to VMs

Hi guys,
I'm planning on buying a Dell poweredge 2900 server. I'll be running
5.2 on it. I plan on using Xen to split that fairly powerful machines
(to us). The thing is, I want to control that certian VMs cannot abuse
the whole machine. Basically, I want to control that a certain VM is for
example limited to 60% of total CPU power, another is limited to 20%,
and another to 20%. In case of no cpu contention, one VM could take up
to 100% of CPU power. Does such a facility exist ?

Also, can (should?) I over commit memory allocations. It's basically the
same problem as partitioning CPU power. The server is gonna have 16Gs.
Do I have to hard partition, or can it be somehow dynamic ? This is more
difficult than CPU partitioning, since if all VMs decide to take max
memory, where would the extra data go :)

Appreciating any pointers, as I haven't used xen in production yet

Thanks, Best Regards

_______________________________________________
rhelv5-list mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/rhelv5-list



_______________________________________________
rhelv5-list mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/rhelv5-list


_______________________________________________
rhelv5-list mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/rhelv5-list


_______________________________________________
rhelv5-list mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/rhelv5-list

Reply via email to