On 2013-01-22, Илья Шипицин wrote:
> I'm running https web reverse proxy.
> at 200-500mbit scale, I see 3500 interrupts per second at em0, em1, also 12
> cpus are running at 70-80%,
> CPU00 is running at "interrupt" level, also there're user processes at
> "user" and "system" levels.
>
>
> under s
> under such load server is experience somewhat to "general network
> delays", network conections become slow (both incoming and outgoing),
> sometimes even 5 sec on 1G network.
It sounds unlikely that CPU congestion is responsible for 5 s network
delays unless your hardware is significantly under
I appreciate your attention for homeopathy and astrology, however I see no
relation of those to CPU00.
Maybe modern processors will handle that stuff, I don't know.
I'm running https web reverse proxy.
at 200-500mbit scale, I see 3500 interrupts per second at em0, em1, also 12
cpus are running at
On Tue, Jan 22, 2013 at 05:37:42PM +0500, ??? wrote:
> I meant OpenBSD feature to use only CPU00 for network things.
> and I am afraid it could cause network issues when some process works on
> CPU00 as well.
OpenBSD is not a real-time OS.
As far as I know there's no intention to make it
I meant OpenBSD feature to use only CPU00 for network things.
and I am afraid it could cause network issues when some process works on
CPU00 as well.
2013/1/22 Gregory Edigarov
> On 01/22/2013 12:55 PM, Gregor Best wrote:
>
>> On Tue, Jan 22, 2013 at 07:56:22PM +1000, David Diggles wrote:
>>
>>
I've seen situations where it has been useful to dedicate a core to a backup
process so the nightly backup would complete, on a busy linux
machine, with a "cpuset".
If this isn't a planned feature in the near future it's not bothering me.
I'm very happy with what OpenBSD does for me.
On Tue, Jan
On 01/22/2013 12:55 PM, Gregor Best wrote:
On Tue, Jan 22, 2013 at 07:56:22PM +1000, David Diggles wrote:
Then if the scheduler always knows what's best, the backup process will be
completely uninhibited, on a system maxed out on all cores.
[...]
What backup process? And why will it be uninhib
On Tue, Jan 22, 2013 at 07:56:22PM +1000, David Diggles wrote:
> Then if the scheduler always knows what's best, the backup process will be
> completely uninhibited, on a system maxed out on all cores.
> [...]
What backup process? And why will it be uninhibited? If the system's
maxed out, all proc
Then if the scheduler always knows what's best, the backup process will be
completely uninhibited, on a system maxed out on all cores.
On Tue, Jan 22, 2013 at 09:29:43AM +0100, Peter Hessler wrote:
> On 2013 Jan 22 (Tue) at 09:25:04 +0500 (+0500), ?? wrote:
> :Hello!
> :
> :I'
On 2013 Jan 22 (Tue) at 09:25:04 +0500 (+0500), Илья Шипицин wrote:
:Hello!
:
:I'm investigating how program should set cpu affinity, is there any
:examples ? (I didn't find any except the "commit that adds cpu affinity
:thing", but there's no user space documentation, no utility, no man page).
:
:
I'm trying to keep CPU00 for network things, and avoid using it for user
applications (there're lots of CPUs).
is it possible to achive it without CPU affinity ?
2013/1/22 Brad Smith
> On Tue, Jan 22, 2013 at 09:25:04AM +0500, ??? wrote:
> > Hello!
> >
> > I'm investigating how program
On Tue, Jan 22, 2013 at 09:25:04AM +0500, ??? wrote:
> Hello!
>
> I'm investigating how program should set cpu affinity, is there any
> examples ? (I didn't find any except the "commit that adds cpu affinity
> thing", but there's no user space documentation, no utility, no man page).
As
Hello!
I'm investigating how program should set cpu affinity, is there any
examples ? (I didn't find any except the "commit that adds cpu affinity
thing", but there's no user space documentation, no utility, no man page).
cheers,
Ilya Shipitsin
13 matches
Mail list logo
