[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [K12OSN] More feedback on Fedora 10 + LTSP



On Tue, Apr 21, 2009 at 4:54 PM, James P. Kinney III
<jkinney localnetsolutions com> wrote:
> The KERNEL is radically different, however. The CentOS version has a
> totally different scheduler than the F10 version. Plus a bazillion other
> changes that have a huge effect on LTSP kernel usage. Due to sound
> changes, it may be challenge to get the CentOS kernel src.rpm and
> compile it for the F10 OS environment but it is certainly worth a try.

I would probably lose what hair I have left if I tried this right now.
 I have 'spare' 8 processor and dual processor Dell servers that the
state was throwing away that I can learn on for this but that won't be
till summer (it is amazing what is sometimes considered obsolete).

> Note also that CentOS kernels are by default tuned more for big
> installation stuff than is the Fedora kernel.

yep, there is a reason that CentOS/RHEL exists.  I would prefer to
stay with CentOS but I have to resolve sound (talk about beating a
dead horse issue) and with LTSP5 I have to admit that sound 'just
works' in all the apps I have to have (except for Audacity but I've
found some documentation that implies it can be made to work).
>
> Again, I strongly suspect you are having a serious kernel issue and
> recommend compiling your own.

It is possible that this is the case. I do know that I have to run
mkinitrd to get the kernel to boot properly since I have an Adaptec
2010s controller. (I20 block device).  There is an outstanding bug wrt
this raid card which will hopefully be resolved 'soon'.  I have to
manually apply a patch/fix.  I'll look at how mkinitrd really works
and see if that patch to make the raid card work isn't causing an
issue. However, I've tested disk I/O and I don't see any issues with
throughput.

> The bridge interface is also a serious slowdown hog. It doesn't seem to
> support gigabit traffic at all. More like 230-400 Mbps. So the idea of
> ditching the bridge and going straight, physical nic is the best,
> fastest route to a speed up. Dig on the Brandon elementary pta site for
> a write up by William Fragakis on how to turn of the bridge process.

I'll look for the notes on how to do this.  The teacher's lesson plan
for the next few sessions involves using OpenOffice/StarOffice and
this works without any issues so I can flip the prior server back into
service if I break anything.  That gives me a weekend in the worst
case to rebuild from scratch.  I'll also say that I've tested
throughput on the external network-facing NIC and it is what I would
expect (it is the same as my systems using CentOS).  I have tested the
bridge though.

> Lastly: you mentioned hyperthreading cpu's. If they are _really_
> hyperthreaded and not full multi-core, turn _off_ the hyperthreading.
> The F10 kernel threading will thrash royally doing stupid task swapping
> as it seems to not understand that the second cpu is a fake one. There
> is a kernel /proc flag to help this but I can't find it right now
> (again).

They are hyperthreaded Xeons.  They have performed very well until
recently. Though, how would I detect thrashing?

> Also be sure to turn off EVERY process not actively used. Some are more
> of a drain on cpu throughput (avahi, a zillion python applets running
> desktop applets for notification - system monitor should be removed!! so
> students won't load it up - it's a HOG)

I've chkconfig'ed to off everything I thought I could but I'll recheck
since I know I didn't turn off avahi.  Is there a list of essential
services somewhere?  Things like bluetooth and the like I know I don't
need on a production system but a nice list would be wonderful.

Sincerely,
Dave Hopkins
Newark Charter School


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]