FAQ Search Today's Posts Mark Forums Read
» Video Reviews

» Linux Archive

Linux-archive is a website aiming to archive linux email lists and to make them easily accessible for linux users/developers.


» Sponsor

» Partners

» Sponsor

Go Back   Linux Archive > CentOS > CentOS

 
 
LinkBack Thread Tools
 
Old 05-10-2010, 12:26 PM
JohnS
 
Default measuring kernel speed

On Sun, 2010-05-09 at 21:46 -0400, Ross Walker wrote:
> On Sat, May 8, 2010 at 7:38 PM, JohnS <jses27@gmail.com> wrote:
> >
> > On Sat, 2010-05-08 at 16:17 -0400, Ross Walker wrote:
> >> On May 8, 2010, at 8:35 AM, Mag Gam <magawake@gmail.com> wrote:
> >>
> >> > At our Physics research labs we do a lot with low latency networks. We
> >> > have been using Centos for over 3 years now and its been great! We
> >> > would like to tune and optimize our setup by removing unneeded
> >> > packages -- kernel modules to be specific. I was wondering, how does
> >> > one measure the speed of the kernel. Is that even possible?
> >>
> >> Use oprofile.
> >>
> >> -Ross
> > ---
> > Ross, never mind I just yummed it onto a machine there faq is inheritly
> > wrong.
>
> The FAQ is only correct in respect to the project's view.
>
> Redhat has a custom oprofile that works with their custom kernels, so
> stock oprofile from the project's site IS incompatible, but that's OK
> cause RH provides one that works with their distro.
>
> -Ross
---
Correct as i found out.

John

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 12:40 PM
Les Mikesell
 
Default measuring kernel speed

JohnS wrote:
> On Sun, 2010-05-09 at 21:46 -0400, Ross Walker wrote:
>> On Sat, May 8, 2010 at 7:38 PM, JohnS <jses27@gmail.com> wrote:
>>> On Sat, 2010-05-08 at 16:17 -0400, Ross Walker wrote:
>>>> On May 8, 2010, at 8:35 AM, Mag Gam <magawake@gmail.com> wrote:
>>>>
>>>>> At our Physics research labs we do a lot with low latency networks. We
>>>>> have been using Centos for over 3 years now and its been great! We
>>>>> would like to tune and optimize our setup by removing unneeded
>>>>> packages -- kernel modules to be specific. I was wondering, how does
>>>>> one measure the speed of the kernel. Is that even possible?
>>>> Use oprofile.
>>>>
>>>> -Ross
>>> ---
>>> Ross, never mind I just yummed it onto a machine there faq is inheritly
>>> wrong.
>> The FAQ is only correct in respect to the project's view.
>>
>> Redhat has a custom oprofile that works with their custom kernels, so
>> stock oprofile from the project's site IS incompatible, but that's OK
>> cause RH provides one that works with their distro.
>>
>> -Ross
> ---
> Correct as i found out.

Would this also be suitable for testing efficiency loss from running under
VMware or other virtualization methods?

--
Les Mikesell
lesmikesell@gmail.com
_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 12:48 PM
JohnS
 
Default measuring kernel speed

On Sun, 2010-05-09 at 11:38 -0400, Mag Gam wrote:
> This is an interesting topic.
>
> So, how does one compare the kernel "speed" from RT and Stock kernel?
>
> Is there a benchmark I can use? For example (I know this is wrong):
> can I look at /proc/cpuinfo and look at the bogmips and compare and
> contrast?
>
---
Please bottom post...or quote inline...

https://rt.wiki.kernel.org/index.php/Main_Page

http://www.latencytop.org/

/proc/cpuinfo and your bogomips is not going to give you what you want
in this respect.

In respect to speed there may different things to measure. Like in the
rt-kernel your looking for the latency of the kernel and improving that.
Which that starts at the raw code layer before it is even compiled. You
will never get good deterministic features from the kernel with out
that. Then there is the filesystem your app is running on.

Like HL7 Messaging the speed facter comes from combining one maching
translating to sending to an upstream provider or sending it to another
host requiring further translation into a database.

Indeed the less kernel module space there is they say it performs faster
but i can't really see that. Maybe someone with more experience will
comment on that for sure. It don't seem to hold true when you allocate
a sole amount of memory to the kernel.

There is a patch required for the mainline kernel to run latencytop.
Oprofile looks promising also that Ross mentioned.

John

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 12:53 PM
JohnS
 
Default measuring kernel speed

On Mon, 2010-05-10 at 07:40 -0500, Les Mikesell wrote:
> JohnS wrote:
> > On Sun, 2010-05-09 at 21:46 -0400, Ross Walker wrote:
> >> On Sat, May 8, 2010 at 7:38 PM, JohnS <jses27@gmail.com> wrote:
> >>> On Sat, 2010-05-08 at 16:17 -0400, Ross Walker wrote:
> >>>> On May 8, 2010, at 8:35 AM, Mag Gam <magawake@gmail.com> wrote:
> >>>>
> >>>>> At our Physics research labs we do a lot with low latency networks. We
> >>>>> have been using Centos for over 3 years now and its been great! We
> >>>>> would like to tune and optimize our setup by removing unneeded
> >>>>> packages -- kernel modules to be specific. I was wondering, how does
> >>>>> one measure the speed of the kernel. Is that even possible?
> >>>> Use oprofile.
> >>>>
> >>>> -Ross
> >>> ---
> >>> Ross, never mind I just yummed it onto a machine there faq is inheritly
> >>> wrong.
> >> The FAQ is only correct in respect to the project's view.
> >>
> >> Redhat has a custom oprofile that works with their custom kernels, so
> >> stock oprofile from the project's site IS incompatible, but that's OK
> >> cause RH provides one that works with their distro.
> >>
> >> -Ross
> > ---
> > Correct as i found out.
>
> Would this also be suitable for testing efficiency loss from running under
> VMware or other virtualization methods?
---
You say efficiency loss. That could mean anything from the power input
down to the kernel. It looks like that can be determined by oprofile
and latencytop. Latencytop will give you the millisecond time for
execution. As far as Oprofile maybe Ross will indeed fill us in if he
can.

John

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 01:46 PM
Ross Walker
 
Default measuring kernel speed

On May 10, 2010, at 8:53 AM, JohnS <jses27@gmail.com> wrote:

>
> On Mon, 2010-05-10 at 07:40 -0500, Les Mikesell wrote:
>> JohnS wrote:
>>> On Sun, 2010-05-09 at 21:46 -0400, Ross Walker wrote:
>>>> On Sat, May 8, 2010 at 7:38 PM, JohnS <jses27@gmail.com> wrote:
>>>>> On Sat, 2010-05-08 at 16:17 -0400, Ross Walker wrote:
>>>>>> On May 8, 2010, at 8:35 AM, Mag Gam <magawake@gmail.com> wrote:
>>>>>>
>>>>>>> At our Physics research labs we do a lot with low latency
>>>>>>> networks. We
>>>>>>> have been using Centos for over 3 years now and its been
>>>>>>> great! We
>>>>>>> would like to tune and optimize our setup by removing unneeded
>>>>>>> packages -- kernel modules to be specific. I was wondering,
>>>>>>> how does
>>>>>>> one measure the speed of the kernel. Is that even possible?
>>>>>> Use oprofile.
>>>>>>
>>>>>> -Ross
>>>>> ---
>>>>> Ross, never mind I just yummed it onto a machine there faq is
>>>>> inheritly
>>>>> wrong.
>>>> The FAQ is only correct in respect to the project's view.
>>>>
>>>> Redhat has a custom oprofile that works with their custom
>>>> kernels, so
>>>> stock oprofile from the project's site IS incompatible, but
>>>> that's OK
>>>> cause RH provides one that works with their distro.
>>>>
>>>> -Ross
>>> ---
>>> Correct as i found out.
>>
>> Would this also be suitable for testing efficiency loss from
>> running under
>> VMware or other virtualization methods?
> ---
> You say efficiency loss. That could mean anything from the power
> input
> down to the kernel. It looks like that can be determined by oprofile
> and latencytop. Latencytop will give you the millisecond time for
> execution. As far as Oprofile maybe Ross will indeed fill us in if he
> can.

Oprofile will show where those precious latencies timings are being
used. It of course adds latency itself, so this should be factored
into the latency timings.

It will time all kernel operations then you can drill down into
particular modules/routines to see more granularity.

Needs debug symbols to be fully useful. Can provide timings as source
code annotations.

This is useful in finding modules, or statically linked routines that
suck up precious time and either eliminate or fix them.

-Ross

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 01:56 PM
Ross Walker
 
Default measuring kernel speed

On May 10, 2010, at 8:40 AM, Les Mikesell <lesmikesell@gmail.com> wrote:

> JohnS wrote:
>> On Sun, 2010-05-09 at 21:46 -0400, Ross Walker wrote:
>>> On Sat, May 8, 2010 at 7:38 PM, JohnS <jses27@gmail.com> wrote:
>>>> On Sat, 2010-05-08 at 16:17 -0400, Ross Walker wrote:
>>>>> On May 8, 2010, at 8:35 AM, Mag Gam <magawake@gmail.com> wrote:
>>>>>
>>>>>> At our Physics research labs we do a lot with low latency
>>>>>> networks. We
>>>>>> have been using Centos for over 3 years now and its been great!
>>>>>> We
>>>>>> would like to tune and optimize our setup by removing unneeded
>>>>>> packages -- kernel modules to be specific. I was wondering, how
>>>>>> does
>>>>>> one measure the speed of the kernel. Is that even possible?
>>>>> Use oprofile.
>>>>>
>>>>> -Ross
>>>> ---
>>>> Ross, never mind I just yummed it onto a machine there faq is
>>>> inheritly
>>>> wrong.
>>> The FAQ is only correct in respect to the project's view.
>>>
>>> Redhat has a custom oprofile that works with their custom kernels,
>>> so
>>> stock oprofile from the project's site IS incompatible, but that's
>>> OK
>>> cause RH provides one that works with their distro.
>>>
>>> -Ross
>> ---
>> Correct as i found out.
>
> Would this also be suitable for testing efficiency loss from running
> under
> VMware or other virtualization methods?

No because oprofile and latencytop's point of reference is just the
running kernel and doesn't factor in CPU allocations, network/disk
virualization/para-virtualization, bandwidth allocations, etc.

Efficiency loss is a slippery slope and VERY configuration dependant.

I have seen VMs perform better than physical machines and I have seen
them perform worse, sometimes on the same physical host!

Go with the "user experience" indicator (assuming it is properly
configured for the workload). Does it seem fast? Then it's fast. Does
it seem slow? Then it is slow.

-Ross

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 02:31 PM
Akemi Yagi
 
Default measuring kernel speed

On Mon, May 10, 2010 at 6:46 AM, Ross Walker <rswwalker@gmail.com> wrote:
> On May 10, 2010, at 8:53 AM, JohnS <jses27@gmail.com> wrote:

>> You say efficiency loss. *That could mean anything from the power
>> input
>> down to the kernel. *It looks like that can be determined by oprofile
>> and latencytop. *Latencytop will give you the millisecond time for
>> execution. As far as Oprofile maybe Ross will indeed fill us in if he
>> can.
>
> Oprofile will show where those precious latencies timings are being
> used. It of course adds latency itself, so this should be factored
> into the latency timings.
>
> It will time all kernel operations then you can drill down into
> particular modules/routines to see more granularity.
>
> Needs debug symbols to be fully useful. Can provide timings as source
> code annotations.

I never thought someone would run oprofile with the RT kernel. I can
upload debuginfo if anyone needs it.

Akemi
_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 02:45 PM
"Brunner, Brian T."
 
Default measuring kernel speed

> -----Original Message-----
> Akemi Yagi [amyagi@gmail.com] wrote
>
> On Sat, May 8, 2010 at 6:49 AM, JohnS <jses27@gmail.com> wrote:
>
> > There is of one place that has a RT Kernel if you want to try it so
> > maybe that person will post a link to this thread for you.
>
> Are you referring to me, John?
>
> You are welcome to provide the link as far as it is stated
> that they are for testing purposes only:
>
> http://centos.toracat.org/kernel/centos5/realtime/

This points to: kernel-rt-2.6.24.7-149.ay.src.rpm et amici
==============================^^^^^^^^
Do you have any kernel-rt-2.6.18-164.15.1.el5. or similarly named that
are the same source base as the current Centos kernel?
==============================^^^^^^^^^^^



************************************************** *****************
This email and any files transmitted with it are confidential and
intended solely for the use of the individual or entity to whom
they are addressed. If you have received this email in error please
notify the system manager. This footnote also confirms that this
email message has been swept for the presence of computer viruses.
www.Hubbell.com - Hubbell Incorporated**

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 03:00 PM
JohnS
 
Default measuring kernel speed

On Mon, 2010-05-10 at 07:31 -0700, Akemi Yagi wrote:
> On Mon, May 10, 2010 at 6:46 AM, Ross Walker <rswwalker@gmail.com> wrote:
> > On May 10, 2010, at 8:53 AM, JohnS <jses27@gmail.com> wrote:
>
> >> You say efficiency loss. That could mean anything from the power
> >> input
> >> down to the kernel. It looks like that can be determined by oprofile
> >> and latencytop. Latencytop will give you the millisecond time for
> >> execution. As far as Oprofile maybe Ross will indeed fill us in if he
> >> can.
> >
> > Oprofile will show where those precious latencies timings are being
> > used. It of course adds latency itself, so this should be factored
> > into the latency timings.
> >
> > It will time all kernel operations then you can drill down into
> > particular modules/routines to see more granularity.
> >
> > Needs debug symbols to be fully useful. Can provide timings as source
> > code annotations.
>
> I never thought someone would run oprofile with the RT kernel. I can
> upload debuginfo if anyone needs it.
>
> Akemi

Actually you need kernel-trace also in there. Trace is what is
recomended by TUV.

John

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 03:10 PM
JohnS
 
Default measuring kernel speed

On Mon, 2010-05-10 at 10:45 -0400, Brunner, Brian T. wrote:
>
> > -----Original Message-----
> > Akemi Yagi [amyagi@gmail.com] wrote
> >
> > On Sat, May 8, 2010 at 6:49 AM, JohnS <jses27@gmail.com> wrote:
> >
> > > There is of one place that has a RT Kernel if you want to try it so
> > > maybe that person will post a link to this thread for you.
> >
> > Are you referring to me, John?
> >
> > You are welcome to provide the link as far as it is stated
> > that they are for testing purposes only:
> >
> > http://centos.toracat.org/kernel/centos5/realtime/
>
> This points to: kernel-rt-2.6.24.7-149.ay.src.rpm et amici
> ==============================^^^^^^^^
> Do you have any kernel-rt-2.6.18-164.15.1.el5. or similarly named that
> are the same source base as the current Centos kernel?
> ==============================^^^^^^^^^^^

kernel-rt-2.6.24.7-149 is the newest Real Time Kernel. RT is based on
2.6.24 and not 2.6.18. So no there is not a 2.6.18-kernel-rt for CentOS
or they ever was one. Akemi just has those for testing.

For as CentOS they it has no RT kernel for it. Not yet, knock knock.

Answer your question?

John

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 

Thread Tools




All times are GMT. The time now is 06:53 AM.

VBulletin, Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO ©2007, Crawlability, Inc.
Copyright 2007 - 2008, www.linux-archive.org