On Thu, Apr 13, 2006 at 10:19:13AM +0900, MAEDA Naoaki wrote:
> Not much for the limit support, but I'm working to improve the accuracy
> of the load calculation. You may know that the load shown in the stats
> file is not so accurate and fluctuating wildly.
>
> In my research, the inaccuracy come
Hi Srivatsa,
Srivatsa Vaddagiri wrote:
On Wed, Feb 22, 2006 at 05:56:31PM +0900, MAEDA Naoaki wrote:
Thank you for your suggestions. I stated to discuss with kurosawa-san
how we can support limit in the current design.
Maeda-san,
I was curious to know if you have made any progress on the lim
On Wed, Feb 22, 2006 at 05:56:31PM +0900, MAEDA Naoaki wrote:
> Thank you for your suggestions. I stated to discuss with kurosawa-san
> how we can support limit in the current design.
Maeda-san,
I was curious to know if you have made any progress on the limit
feature ..
--
Regards,
vats
On Wed, 2006-02-22 at 15:03 +0530, Srivatsa Vaddagiri wrote:
> On Tue, Feb 21, 2006 at 07:24:15AM -0800, Chandra Seetharaman wrote:
> > I think Vatsa is differentiating between "hard limit"(class will not get
> > any resource and the surplus resources will be wasted) and "soft
> > limit" (class can
On Tue, Feb 21, 2006 at 07:24:15AM -0800, Chandra Seetharaman wrote:
> I think Vatsa is differentiating between "hard limit"(class will not get
> any resource and the surplus resources will be wasted) and "soft
> limit" (class can get resource over its soft limit if there is surplus).
Yes, this is
Hi,
> > > > To support the limit, surplus CPU is waisted. I don't think
> > > > it is a good design. In other words, if every class is satisfied
> > > > in terms of its CPU guarantee and some classes can consume more,
> > > > surplus CPU should be given to the classes. Therefore, I'm not
> > > >
On Tue, 2006-02-21 at 15:59 +0900, MAEDA Naoaki wrote:
> Hi vatsa,
>
> > > To support the limit, surplus CPU is waisted. I don't think
> > > it is a good design. In other words, if every class is satisfied
> > > in terms of its CPU guarantee and some classes can consume more,
> > > surplus CPU sh
Hi vatsa,
> > To support the limit, surplus CPU is waisted. I don't think
> > it is a good design. In other words, if every class is satisfied
> > in terms of its CPU guarantee and some classes can consume more,
> > surplus CPU should be given to the classes. Therefore, I'm not
> > interested in
On Tue, Feb 21, 2006 at 11:24:43AM +0900, MAEDA Naoaki wrote:
> To support the limit, surplus CPU is waisted. I don't think
> it is a good design. In other words, if every class is satisfied
> in terms of its CPU guarantee and some classes can consume more,
> surplus CPU should be given to the cla
Hi,
> Thanks a lot Maeda.
> I tried that and it worked !
> I gave 20% share to kernbench and it showed 22% in the output. Please
> see the attachment. I plan to play more with it.
That's great.
> Can you pls explain why limit is meaningless for the CPU resource
> controller..?
To support the l
Thanks a lot Maeda.
I tried that and it worked !
I gave 20% share to kernbench and it showed 22% in the output. Please
see the attachment. I plan to play more with it.
Can you pls explain why limit is meaningless for the CPU resource
controller..?
Regards,
Rajaram Suryanarayanan | Team Lead | ES7
Hi again,
> Thanks. So is it enough to run a while(1) process in the default class
> to see a difference ?
Yes. It's enough.
> Also do you mean I am correct in my methods of configuring shares and
> using the kernbench tool ?
I think so.
> Now I feel that I should have moved the pid of kernb
Hi Maeda,
>
> I find that Percent CPU is 98.5 in the first file and 97.5 in the
second file. Just 1% decrease for configuring 20% share of the CPU ?
Another CPU hog process, such as an infinite loop must be run on the
other
class. Otherwise kernbench can consume whole CPU time regardles
Hi Rajaram,
> Hi,
>
> I am trying to use the kernbench tool for trying out the difference the CPU
> resource controller makes in terms of allocating CPU time.
>
> I am using a one CPU machine with nearly 0.99 GB RAM.
>
> I ran kernbench tool twice and compared the results.
>
> First time, wit
14 matches
Mail list logo
