Re: compiling 2.6

From: William D. Tallman (wtallman_at_olypen.com)
Date: 03/18/04 

Date: Wed, 17 Mar 2004 19:11:17 -0500

P.T. Breuer wrote:

> William D. Tallman <wtallman@olypen.com> wrote:
>> I've gathered that compiling all the functionality into the kernel only
>> make sense for servers, where the function of the machine itself is well
>
> It doesn't make sense there, since you can configure by loading modules
> just as well as you can configure by linking at compile time instead!

You're right, it doesn't make complete sense to me either. If everything is
loaded at boot, then one would think that would be equivalent to having it
all compiled in directly. But I'm not certain that's true. I really have
to get down and study the kernel at some point, but I've got a lot of other
things in the queue ahead of that... <sigh>

The question is, I guess, whether or not there is a difference in this
regard. Servers are typically evaluated in terms of response time
increments that are meaningless to desktops and work stations, or so I
understand. So I gather that the question has intrinsic merit, whatever
the case.
 
>> defined. That makes for much faster response, of course.
>
> Where do you get this from??? There's no significant response
> difference that I can think of! It's a question of when the linking is
> done, and what mode it takes, that's all. There would certainly
> usually be an extra indirection to resolve in an inter-module function
> lookup, but so? If you worry about that you should file your
> fingernails down so you can type faster!

Okay, I'm hearing this from guys who run service centers. I don't know the
particulars, as I'm not interested in running a server (that's always
subject to change, though... <grin>). I took their word for this; hence,
the "of course". I figure they know whereof they speak.

Clearly it makes no discernable difference for home boxes and workstations,
which is why I always go modules, even though some of them are loaded at
boot. It's easier to add a module than to recompile the kernel itself.
 
Make sense?

Bill Tallman 

-- 
Registered Linux User: #221586
Mdk-10.0 and Slackware-9.1
A Luxuriance of Linux!!!

Relevant Pages

  • Re: Power-off on shutdown -h
    ... Make sure you have support for power management compiled into your kernel. ... here) which means you'll have to do some compiling if you ... though they are linux servers. ...
    (RedHat)
  • Re: compiling 2.6
    ... > I've gathered that compiling all the functionality into the kernel only make ... > sense for servers, where the function of the machine itself is well ... It's a question of when the linking is ...
    (comp.os.linux.misc)
  • FreeBSD 6.1, crashes and a lack of vmcores
    ... These servers run under extremely high load through a majority of the day and run a mix of postfix, MySQL and custom filter software using MFS partitions. ... The biggest frustration in this is that of the few dozen crashes we've had I've only been able to get one successful dump. ... kernel: kernel dumps on /dev/ad0s1b ...
    (freebsd-hackers)
  • Re: Kernel does strange things when compilations push memory usage above physical memory and the com
    ... oom-killer appears to have killed stuff. ... Does the kernel need to be compiled with any special options to have ... I have what might be a similar issue, after a lot of compiling on a ramdisk. ... but alternative responses that the system has made to such ...
    (Linux-Kernel)
  • Re: Kernel does strange things when compilations push memory usage above physical memory and the com
    ... I can confirm that this issue also affects kernel 2.6.33.5. ... oom-killer appears to have killed stuff. ... I have what might be a similar issue, after a lot of compiling on a ramdisk. ... but alternative responses that the system has made to such ...
    (Linux-Kernel)