You are not logged in.
Yeah, this question has been with me for quite some time. The kernel26 Archlinux ships is built for a low-latency desktop with high timer interrupts and preempt. For my server this is just unneccesary. But, having to recompile it every time is a pita, since I don't really change anything except for the processor type (core2), latency timer (100Hz) and disabling preempt.
Now I was wondering if there are enough users that would like this kernel to be a part of the Archlinux repos. Not that I'm offering to be a maintainer (oh i feel so lame by saying this... ) but asking this to see how you guys feel about it.
We could then have a kernel26-server package with qemu-server and kqemu-server and so. But this would only be useful if enough users are with me on this one.
So make yourself known!
Last edited by voidzero (2009-04-22 14:23:30)
Offline
Is there no one interested? Or did I just post this at a wrong point in time
Offline
i'm all for it, but unless the changes have measurable benefits, it's rather a moot exercise. Have you measured the effects between the stock kernel and your server config?
Offline
I'm for it also, i think this would be a good move for the devs and "hopefully" get more people to use arch on a server..
Offline
i thought using arch on a sever was a no go due to its state of the art rolling release nature and bleeding edge software updates....
maybe there should be a new repo ala testing called "safe" which includes outdated tried and true versions of packages... but it might prove to be too much work.
Offline
To be honest I haven't tested it, but it feels a lot more responsive altogether. As to the 'safe' repo, well I just have testing way down the line. It would be a good idea to have a lesser updated repo, but in general I haven't had a problem with using core as the first repo.
Offline
I'm interested of a new kernel if it includes HIGHMEM64G=y in its config file. There are plenty of users with more than 3Gigs of RAM unwilling to switch to 64bit for various reasons. It's said HIGHMEM64G has little performance impact, but with newer processors (Pentium4 and newer) I was unable to see any relevant performance impact. I've tested it on Slackware and Gentoo installations over time and it works flawlessly (even with proprietary drivers such as nvidia's and amd's).
In love I believe and in Linux I trust
Offline
If you want the Arch devs to do this, submit a feature request, but don't expect an overwhelmingly positive response. A much better idea would be for some or all of those interested to initiate a community project to provide a server-optimised kernel and associated packages.
Offline
I'll try to avoid this situation. Anyway for my servers I almost always compile custom kernels. I was speaking for the desktop users about the 4G limitation (3Gigs usable + 1G adress space). The memory modules are really cheap right now, and many of us have desktops with 4GB. And I don't wanna switch some of my systems to Arch64 because of the lack of packages needed from time to time (like wine).
In love I believe and in Linux I trust
Offline