Hi,

I ask this question out of curiosity, I used to re-compile custom kernel's many years ago on multiple versions of Linux and I've not done it for a long time.

I've just got back into Linux (shame on me, I've gone for Debian though I would have gone for Redhat had it not been enterprise and yes I know I could have used CentOS) and now have a working LAMP/Mail/"Anything I want it to do" server which I have thoroughly enjoyed setting up.

Is it worth doing and relearning? Don't get me wrong I used to enjoy it.

Thanks,

Scoot

If you are looking for more speed, not worth the time. If you want to learn, learning is fun and I say go for it.

I used to compile my own back in the day. It has however come to a point where the returns are about none. The speed of hardware is just insane nowadays. I did a Gentoo on this laptop about 8 months ago just to check. I was in front of a full compiled XFCE desktop in about 3 hours. Insane (I did it a long time ago and it took days).

My experience has been that a custom kernel can significantly speed things up and provide better performance, provided you pick the correct kernel configuration options prior to running make.

My experience in the past as a system administrator has been that the default kernel is used in a production environment. This guarantees stability, kernel security updates, and support form the distribution maintainer. It is generally better to use a well tested kernel in a server environment; the distribution kernel is just that.

For personal use, I usually build a custom kernel using the newest long term support (LTS) kernel available. Then every few releases of that LTS kernel, I recompile for bug fixes and other updates. I usually leave the distribution kernel installed on servers though.

I am positive others will have different outlooks/opinions. For personal use, it really is a matter of preference. The exceptions are if you are trying to optimize, you have a very specific hardware setup across multiple machines, or you are trying to reach a policy specification for specific use cases (an example would be for security/ system hardening).

The majority of the time, the distribution provided kernel is the best option. This is because it is tested and it will most likely run well, regardless of whatever you throw at it.

On a home system it would end up being a learning project for the most part.

On a production server it could improve your server use to some degree. You'd have to set any number of the settings for your load. You'd have to remove all the hundreds of extra drivers too. As to how far it would improve your system would be the real test. I'd think a well run server farm would have their own kernel. Millions of servers run stock kernels. It can be considered a safer move.

Anyone interested in linux ought to do it once I'd think. It can be rather time consuming the first time.

to LEARN to do it
-- YES --
and who knows you might need to build a custom kernel for very odd and unusual hardware

but as above it is a very good learning experience
like building a SECOND install of say glib for a development version of some software
-- The Gimp DEVELOPMENT branch is notorious for things like that

for the EXPERIENCE go for it

to speed up booting and the OS or to SHRINK the size ????

Not worth it too much , but there is always an exception to the rule



to quote above
i do this with the image editor "The Gimp "
after 14 or so years of building the source , this is just a habit i do

If you want to, do it.
I've been using Linux & BSD since 1999, & only twice have I compiled a kernel, & that was just for curiosity.

What is point of calling the system Linux and not compile the kernel for your specific taste or need. I have done it many a times upto now. You have to learn a lot more for building your own kernel and learning is needed constantly as features and options change with versions. Though some options remain same.

Thanks for the replies I think I'll do it on my main pc, I don't really want to have to recomplile on the server every time a new one comes out for security reasons. Quite looking forward to seeing what's in there these days.

Absent a specific hardware need to do so, I would think that it doesn't make sense to compile a custom kernel. When you do, you invite all kinds of compatibility issues with other software packages.