You are not logged in.
Pages: 1
Relatively new arch user here, so feel free to flame me for RTFM. I have been reading the docs, but haven't stumbled across anything related to this issue.
I'm going to be installing arch on servers and desktops, and noticed there is only one kernel package, presumably optimized for desktop users. I'm wondering if this means that I need to configure a different kernel package for my server installs? For example, should I be be using a different default scheduler (i.e. deadline instead of cfq), or, maybe the kernel devs have already resolved this at runtime.
Also note the comment that I've heard frequently "don't run arch on production servers" doesn't apply. Save for critical security updates, we already stage all deployments (even using Ubuntu) on testing systems for at least a month before deploying to production. I expect the arch servers will be just as stable as anything else out there (probably more, actually).
Last edited by pgoetz (2014-02-22 10:52:28)
Offline
You don't need a different kernel, but you might want one. I run my extremely non-critical home server on the stock kernel and it works just fine.
Re the single example you gave - yes, you can select your i/o scheduler at runtime.
Offline
The select scheduler at runtime is a good example. I'm guessing that everything in the kernel so so modular these days that this might be a non-issue; I'm just fishing for some deeper insights than I might have on this issue.
Offline
We use the core/linux kernel on our servers. We even use the default CFQ IO scheduler, since these days, it is pretty good for most workloads.
Offline
Some kernel patches to consider:
BFS for desktop responsiveness (here's the patch I've put together for kernel 3.10)
PHC for CPU undervolting - cooler is always better
BFQ disk scheduler
Full AppArmor support
Offline
Pages: 1