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, 03:11 PM
Akemi Yagi
 
Default measuring kernel speed

On Mon, May 10, 2010 at 8:00 AM, JohnS <jses27@gmail.com> wrote:
>
> 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:

>> > 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.

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

I have kernel-rt-trace up there ...

Akemi
_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 03:17 PM
"Brunner, Brian T."
 
Default measuring kernel speed

> -----Original Message-----
>
> 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 nor ever was one. Akemi just has those for testing.
>
> For as CentOS it has no RT kernel. Not yet, knock knock.
>
> Answer your question?
>
> John

It answered that question (thanks!) and uncovered another:

Has anybody run CentOS 5 with the rt kernel that Akemi Yagi has built?

If so, what modules etc have to be updated to use it? (this is asking,
has anybody mapped the minefield?)
************************************************** *****************
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, 03:28 PM
Akemi Yagi
 
Default measuring kernel speed

On Mon, May 10, 2010 at 8:17 AM, Brunner, Brian T.
<BBrunner@gai-tronics.com> wrote:

> It answered that question (thanks!) and uncovered another:
>
> Has anybody run CentOS 5 with the rt kernel that Akemi Yagi has built?
>
> If so, what modules etc have to be updated to use it? (this is asking,
> has anybody mapped the minefield?)

I (naturally) test-installed kernel-rt and ran it for a while. But I
did not do anything in particular.

By the way, debuginfo is being uploaded. It takes a while because they
are large ...

Akemi
_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 03:45 PM
JohnS
 
Default measuring kernel speed

On Mon, 2010-05-10 at 11:17 -0400, Brunner, Brian T. wrote:
>
> > -----Original Message-----
> >
> > 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 nor ever was one. Akemi just has those for testing.
> >
> > For as CentOS it has no RT kernel. Not yet, knock knock.
> >
> > Answer your question?
> >
> > John
>
> It answered that question (thanks!) and uncovered another:
>
> Has anybody run CentOS 5 with the rt kernel that Akemi Yagi has built?
>
> If so, what modules etc have to be updated to use it? (this is asking,
> has anybody mapped the minefield?)
---
Yes I can say there is nothing wrong with Akemis. I have ran them.
They did not explode. The ones I have built have not exploded either
and the only down time is for reboots and hardware firmware updates on
the servers.

I can positively say I have had no probs on Dell and IBM even my own
desktop and various HP desktops.

Kmods are a no go at present so forget those.

They run very suprisingly and responsive under heavy cpu and memory
bound tasks. Even when the interrupt service is off and done manually
is really when the performance come in and assigning single task per
cpu. It's no kernel unless it's on a smp machine.

All testing and running has been under CentOS 5.4s current gcc building.

John

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 03:52 PM
Les Mikesell
 
Default measuring kernel speed

On 5/10/2010 8:56 AM, Ross Walker wrote:
>
>>
>> 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.

Realistically, VM performance is going to depend mostly on how much
contention you have between guests for common resources especially if
you overcommit them. But, I'd like to have some idea of how much effect
running under VMware ESXi would have for a single guest, compared to
running directly on the hardware. If there's not a big loss (and it
doesn't 'feel' like there is), I'd consider this worthwhile for servers
doing oddball things where its not worth the trouble to script a
re-install for every little app someone might have running as a means to
deal with the usual pain of moving a working system to different
hardware. Plus, if there is extra capacity you can bring up another
virtual machine or test the next version almost for free, and you get an
almost-hardware level kvm too (after the base install works and you have
an IP address...). I'd just like to have a more objective measure of
what it costs in performance.

--
Les Mikesell
lesmikesell@gmail.com


_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 03:57 PM
JohnS
 
Default measuring kernel speed

On Mon, 2010-05-10 at 11:17 -0400, Brunner, Brian T. wrote:
>
> > -----Original Message-----
> >
> > 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 nor ever was one. Akemi just has those for testing.
> >
> > For as CentOS it has no RT kernel. Not yet, knock knock.
> >
> > Answer your question?
> >
> > John
>
> It answered that question (thanks!) and uncovered another:
>
> Has anybody run CentOS 5 with the rt kernel that Akemi Yagi has built?
>
> If so, what modules etc have to be updated to use it? (this is asking,
> has anybody mapped the minefield?)
---
One last thing:

They are not for Desktop User machines. Unless they are part of the MRG
messaging and grid configured to use spare cpu cycles.

John

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 04:37 PM
Ross Walker
 
Default measuring kernel speed

On Mon, May 10, 2010 at 11:52 AM, Les Mikesell <lesmikesell@gmail.com> wrote:
> On 5/10/2010 8:56 AM, Ross Walker wrote:
>>
>>>
>>> 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.
>
> Realistically, VM performance is going to depend mostly on how much
> contention you have between guests for common resources especially if
> you overcommit them. *But, I'd like to have some idea of how much effect
> running under VMware ESXi would have for a single guest, compared to
> running directly on the hardware. *If there's not a big loss (and it
> doesn't 'feel' like there is), I'd consider this worthwhile for servers
> doing oddball things where its not worth the trouble to script a
> re-install for every little app someone might have running as a means to
> deal with the usual pain of moving a working system to different
> hardware. *Plus, if there is extra capacity you can bring up another
> virtual machine or test the next version almost for free, and you get an
> almost-hardware level kvm too (after the base install works and you have
> an IP address...). *I'd just like to have a more objective measure of
> what it costs in performance.

With ESXi you can really control the contention on resources with
allocation policies, so applications that really need the resources
get them.

As with any system these days the biggest contention is going to be
disk and network. Make sure storage is setup appropriately for the
application, just cause the server is virtual doesn't mean you can
lump all the application's data onto one common datastore, keep a
datastore for the OS (which can be shared for all VMs) and a separate
iSCSI/Fiber datastore/lun for each application's data. You can use
RDMs if your OS is on a VMFS datastore, or do iSCSI directly in the VM
if you use NFS datastores for the OS.

You will notice minimal degradation running a single VM under ESXi.

I have ESXi hosts here running 20 VMs per host with some doing
terminal services, some doing email, some doing database and other
network services and I have not noticed any diminished performance,
and yes going virtual is simply the easiest way to perform upgrades.

-Ross
_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 05:15 PM
Les Mikesell
 
Default measuring kernel speed

On 5/10/2010 11:37 AM, Ross Walker wrote:
>
> I have ESXi hosts here running 20 VMs per host with some doing
> terminal services, some doing email, some doing database and other
> network services and I have not noticed any diminished performance,
> and yes going virtual is simply the easiest way to perform upgrades.

I think it is unfortunate how difficult it is to back up a working linux
machine and restore it onto different hardware, given that the system
really is very hardware independent. But, detecting the hardware and
mapping it to device drivers seems to be a black art hidden inside of
anaconda and then the local hardware related settings are fairly
hopelessly intertwined with application and user preferences in your
backup copies. I always thought that this would be a common enough
problem that some distribution would address it, but so far it hasn't
happened.

--
Les Mikesell
lesmikesell@gmail.com


_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 05:37 PM
Ross Walker
 
Default measuring kernel speed

On Mon, May 10, 2010 at 1:15 PM, Les Mikesell <lesmikesell@gmail.com> wrote:
> On 5/10/2010 11:37 AM, Ross Walker wrote:
>>
>> I have ESXi hosts here running 20 VMs per host with some doing
>> terminal services, some doing email, some doing database and other
>> network services and I have not noticed any diminished performance,
>> and yes going virtual is simply the easiest way to perform upgrades.
>
> I think it is unfortunate how difficult it is to back up a working linux
> machine and restore it onto different hardware, given that the system
> really is very hardware independent. *But, detecting the hardware and
> mapping it to device drivers seems to be a black art hidden inside of
> anaconda and then the local hardware related settings are fairly
> hopelessly intertwined with application and user preferences in your
> backup copies. *I always thought that this would be a common enough
> problem that some distribution would address it, but so far it hasn't
> happened.

That's why God invented the systems administrator!

Here I use kickstart scripts for baseline server types that perform
all the basic configurations on install. Then I typically keep the
server-centric config in a common location, /etc/<servername> and use
symbolic links to the system supplied config, this can also be
scripted for quick recovery. I keep the application data on separate
volumes then the OS (typical OS image is 8GB, most is swap) using
iSCSI so I can connect to them from another VM easily enough (RDM or
direct iSCSI), everything is installed via RPM, if the distro repo
version isn't adequate I build my own and keep a custom in-house repo,
no third party repos.

I have yet to look at Cobbler which is suppose to simplify the
creation and management of all these kickstart scripts and provide a
nice interface, but haven't had the time yet.

If I am setting up an ESXi infrastructure the first thing I would do
is setup a Cobbler server and a Windows deployment server (maybe a
Solaris Jump Start server) and integrate it with the VMware vCenter
templates. Then it's all point-n-click server deployment from there.

-Ross
_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 05-10-2010, 06:05 PM
Les Mikesell
 
Default measuring kernel speed

On 5/10/2010 12:37 PM, Ross Walker wrote:
>
>> I think it is unfortunate how difficult it is to back up a working linux
>> machine and restore it onto different hardware, given that the system
>> really is very hardware independent. But, detecting the hardware and
>> mapping it to device drivers seems to be a black art hidden inside of
>> anaconda and then the local hardware related settings are fairly
>> hopelessly intertwined with application and user preferences in your
>> backup copies. I always thought that this would be a common enough
>> problem that some distribution would address it, but so far it hasn't
>> happened.
>
> That's why God invented the systems administrator!

Well, yeah - I suppose you could say the design is good for the job
security of sysadmins and for requiring support subscriptions from the
distribution vendors, but it's something that the computer really should
be able to handle by itself just like it does during the initial install.

> Here I use kickstart scripts for baseline server types that perform
> all the basic configurations on install.

Have you totaled up the hours you've spend on these tasks that would be
unnecessary in a better-designed system? And even if that sort-of makes
sense for servers that have a basic "type", what about ones that have
application developers as users and end up accumulating all kinds of
cruft that you don't know about?

> If I am setting up an ESXi infrastructure the first thing I would do
> is setup a Cobbler server and a Windows deployment server (maybe a
> Solaris Jump Start server) and integrate it with the VMware vCenter
> templates. Then it's all point-n-click server deployment from there.

Don't forget that you can use ESXi for free, but not vCenter. But,
there's really no problem in just copying/cloning VMware images around -
you don't have to go through any extra contortions to be able to
reproduce them (with variations for every OS), you just need to save a
baseline copy before adding specialized applications.

--
Les Mikesell
lesmikesell@gmail.com



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

Thread Tools




All times are GMT. The time now is 03:59 PM.

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