|
|
|
|
|
|
|
|
|
|
xen-users
Re: [Xen-users] CPU load balancing
I'm not sure if the idea you are proposing is really useful. The
credit scheduler will already allocate unused CPU time to other VMs,
if you don't have a cap configured, it does not necessarily follow
weight percentage allocation exactly.
You might find this research paper helpful:
http://www.cs.ucsd.edu/~dgupta/papers/per07-3sched-xen.pdf
On Wed, Jul 23, 2008 at 5:39 AM, Andreï V. Fomitchev <fomitchev@xxxxxxxx> wrote:
> Hi all.
>
> I continue to study the XEN possibilities and currently I search a 'CPU load
> balancing tool' which answers the scenario below:
> Configuration: using sched-credit command, the 4 DomU are the same weight
> and each uses only 15% of CPU.
> The balancing tool monitors the DomU CPU loads and if two (or more)
> consecutive measures of CPU load equal to 100% of CPU and if the other
> domains are in Idle mode, the tool increases the Domain weight and allocates
> more of CPU. When the Domain terminated and "returns" to Idle, the tool
> returns the Domain configuration in nominal mode.
>
> 1) Because I did not find a response in XEN documentation, my first
> questions are "philosophical":
> - this kind of tools is it interesting? Can it provide some improvements of
> performances on-the-fly?
>
> 2) I found some tools which manage the load balancing of entire clusters. I
> think that it is possible to adapt them or their algorithm to CPU balancing
> but I think also that I am not the first to meet this issue... Did you read
> something about a tool which can answer my needs?
>
> Thank you for your answers.
>
> Best regards,
>
> --
> Andreï V. FOMITCHEV
>
> ________________________________
> Envoyé avec Yahoo! Mail.
> Une boite mail plus intelligente.
> _______________________________________________
> Xen-users mailing list
> Xen-users@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-users
>
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users
|
|
|
|
|