Basic questions on the kernel

Basic questions on the kernel

[Girish Kale]

Hi,

I have a few basic questions on the kernel performance:

1) For embedded systems, even though standard linux may not have hard
real-time performance unless real-time patch, low latency patch are
applied, I would like to know if linux can give "deterministic"
performance. Meaning can we be guaranteed that a particular process will
be scheduled after a particular time. What are the these times or where
do I find this data.  

2) When we are using linux on an embedded system, will there a memory
management unit for this, since there will not be any swapping in/out of
pages? Can we avoid a process from keeping on allocating memory thereby
starving other applications? 

Any pointers to the above topics will be appreciated.

Thanks,
Girish
-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: Basic questions on the kernel

[Burton Samograd]

"Girish Kale" <girish.kale@nevisnetworks.com> writes:
> I have a few basic questions on the kernel performance:
> 
> 1) For embedded systems, even though standard linux may not have hard
> real-time performance unless real-time patch, low latency patch are
> applied, I would like to know if linux can give "deterministic"
> performance. Meaning can we be guaranteed that a particular process will
> be scheduled after a particular time. What are the these times or where
> do I find this data.

There is no performance "guarantee" for any version of linux kernel
other than the branches which are geared toward real time use (such as
RTLinux).  

> 2) When we are using linux on an embedded system, will there a memory
> management unit for this, since there will not be any swapping in/out of
> pages? Can we avoid a process from keeping on allocating memory thereby
> starving other applications?

You still need a memory managment unit for handling protection between
running processes. You might not need swap on your system, but you'll
need that.

As an aside, I've never been a fan of using something like linux (or
any unix for that matter) for embedded systems.  There are plenty of
other real-time operating systems which do the job better in a smaller
space, and a lot of times even lowly DOS is good enough for the job
(and it's even decent soft real-time as well).

> Any pointers to the above topics will be appreciated.

Try searching for 'embedded linux' on google and you'll find some more
good information I'm sure.

-- 
burton samograd
kruhft@kruhft.dyndns.org
http://kruhftwerk.dyndns.org

-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: Basic questions on the kernel

[Stephen Samuel]

Girish Kale wrote:
> Hi,
> 
> I have a few basic questions on the kernel performance:
> 
> 1) For embedded systems, even though standard linux may not have hard
> real-time performance unless real-time patch, low latency patch are
> applied, I would like to know if linux can give "deterministic"
> performance. Meaning can we be guaranteed that a particular process will
> be scheduled after a particular time. What are the these times or where
> do I find this data.  
You can only get "deterministic" performance if you're using a real-time
patch. I'm not sure what the timing is, but I believe that it's in the
sub-milisecond range.

You might be able to get near-realtime (but no absolute guarantees)
by setting the process that needs fast response to nice -20 this
gives it highest priority, so that it gets CPU pretty much whenever
it wants it (after device drivers).


> 
> 2) When we are using linux on an embedded system, will there a memory
> management unit for this, since there will not be any swapping in/out of
> pages? Can we avoid a process from keeping on allocating memory thereby
> starving other applications? 

Linux generally wants an MMU. Embeded kernels MAY be able to
survive without one (dunno).

Check the setrlimit call, or the ulimit bash built=in for info on
how to set resource usage limits.

You can also set limits on a system-wide basis by using the /proc mechanisms.
(your system probably has a formalized way of doing that automatically on boot..
for RedHat, that would be the file /etc/sysctl.conf )



-- 
Stephen Samuel +1(604)876-0426                samuel@bcgreen.com
		   http://www.bcgreen.com/~samuel/
Powerful committed communication, reaching through fear, uncertainty and
doubt to touch the jewel within each person and bring it to life.

-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

RE: Basic questions on the kernel

[Paul Kraus]

I unsubbed from this list a month ago and I still get messages. Not as
many but I still get them. 

-----Original Message-----
From: linux-newbie-owner@vger.kernel.org
[mailto:linux-newbie-owner@vger.kernel.org] On Behalf Of Stephen Samuel
Sent: Friday, May 30, 2003 1:17 PM
To: Girish Kale; linux-newbie@vger.kernel.org
Subject: Re: Basic questions on the kernel


Girish Kale wrote:
> Hi,
> 
> I have a few basic questions on the kernel performance:
> 
> 1) For embedded systems, even though standard linux may not have hard 
> real-time performance unless real-time patch, low latency patch are 
> applied, I would like to know if linux can give "deterministic" 
> performance. Meaning can we be guaranteed that a particular process 
> will be scheduled after a particular time. What are the these times or

> where do I find this data.
You can only get "deterministic" performance if you're using a real-time
patch. I'm not sure what the timing is, but I believe that it's in the
sub-milisecond range.

You might be able to get near-realtime (but no absolute guarantees) by
setting the process that needs fast response to nice -20 this gives it
highest priority, so that it gets CPU pretty much whenever it wants it
(after device drivers).


> 
> 2) When we are using linux on an embedded system, will there a memory 
> management unit for this, since there will not be any swapping in/out 
> of pages? Can we avoid a process from keeping on allocating memory 
> thereby starving other applications?

Linux generally wants an MMU. Embeded kernels MAY be able to survive
without one (dunno).

Check the setrlimit call, or the ulimit bash built=in for info on how to
set resource usage limits.

You can also set limits on a system-wide basis by using the /proc
mechanisms. (your system probably has a formalized way of doing that
automatically on boot.. for RedHat, that would be the file
/etc/sysctl.conf )



-- 
Stephen Samuel +1(604)876-0426                samuel@bcgreen.com
		   http://www.bcgreen.com/~samuel/
Powerful committed communication, reaching through fear, uncertainty and
doubt to touch the jewel within each person and bring it to life.

-
To unsubscribe from this list: send the line "unsubscribe linux-newbie"
in the body of a message to majordomo@vger.kernel.org More majordomo
info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

RE: Basic questions on the kernel

[Ray Olszewski]

At 02:17 PM 5/30/2003 -0400, Paul Kraus wrote:
>I unsubbed from this list a month ago and I still get messages. Not as
>many but I still get them.

I'm sorry this is happening to you, Paul. But as you probably remember from 
when you were subscribed and asking questions here, no one actually *on* 
this list is its administrator, so nobody who is seeing this complaint can 
fix the underlying problem (whatever it might be) that is causing the 
annoyance.

The customary majordomo address for a list administrator is list_name-owner 
(in this instance, linux-newbie-owner@vger.kernel.org). I suggest you write 
there ... though this list is, to my knowledge, pretty much ignored by 
whoever its admin is, so you may not get any help that way either ... but I 
don't know of any alternative, except perhaps autobouncing the messages 
(preferably in a way that gets the admin's attention but does not come back 
to the list) or spam-filtering them without bouncing.

Note that I'm cc'ing you on this reply directly, so you *should* get it (at 
least once).

>-----Original Message-----
>From: linux-newbie-owner@vger.kernel.org
>[mailto:linux-newbie-owner@vger.kernel.org] On Behalf Of Stephen Samuel
>Sent: Friday, May 30, 2003 1:17 PM
>To: Girish Kale; linux-newbie@vger.kernel.org
>Subject: Re: Basic questions on the kernel
[old stuff deleted]



-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: Basic questions on the kernel

[oford]

[-- Attachment #1: Type: text/plain, Size: 1403 bytes --]

On Fri, 2003-05-30 at 08:00, Girish Kale wrote:
> Hi,
> 
> I have a few basic questions on the kernel performance:
> 
> 1) For embedded systems, even though standard linux may not have hard
> real-time performance unless real-time patch, low latency patch are
> applied, I would like to know if linux can give "deterministic"
> performance. Meaning can we be guaranteed that a particular process will
> be scheduled after a particular time. What are the these times or where
> do I find this data.  
> 
> 2) When we are using linux on an embedded system, will there a memory
> management unit for this, since there will not be any swapping in/out of
> pages? Can we avoid a process from keeping on allocating memory thereby
> starving other applications? 

1) I think your best source of info here is going to be the scheduler
code itself.  You can find lots of information on it by searching the
linux-kernel mailing list archives.  Here is the link:
http://marc.theaimsgroup.com/?l=linux-kernel&r=1&w=2

2) There are MMUs for almost every architecture.  My advice is to email
the kernel maintainer of the architecture you wish to use.  Seek out any
lists that the embedded community uses and I'm sure they will be glad to
assist you.

3) Be prepared to make modifications to the kernel source code to fit
the needs of your specific device.
-- 
Owen Ford <oford@ev1.net>

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]