Steve Wolfe wrote:
>
> > > RPM kernel upgrades have always worked fine for me, just do an
> "rpm -i"
> > > (as opposed to -U, so you keep your old kernel around for good
> measure),
> > > then set up lilo to point to the new kernel image.
>
> At the risk of starting a flame-war, I avoid RedHat's "standard" kernels
> like the plague, as my network throughput is significantly slower. My
> compiling a kernel with exactly what I need, I get a much higher
> throughput - along the order of several megabytes per second. With
> everything that they enable in their kernels, I imagine that other aspects
> would also suffer. Whether that extra bit of performance is necessary
> depends, of course, on your situation. YMMV.
>
> steve
I usually avoid kernel rpms also, but not necessarily because I dislike
RH...they just don't keep up versions and features I'm interested in. On
the other hand, I have currently the 2.4.6-pre1-xfs kernel, non-RH, and
the eepro100 throughput is absolutely miserable. A while back, on
earlier kernels (2.2.x), I was able to copy the entire X11 distribution
(unpacked, untarred source) within a couple of seconds over the
ethernet. I recently tried to copy kernel source the same way, and after
many many hours (I watched a couple of two hour movies on television),
gave up, it still was not finished. There have been other people talking
about the issue on the kernel list, so I suspect this is fixed on newer
kernels, but I have not yet got around to testing (lack of reasonable
Internet connection in my area means downloading kernels and anything
over a couple MB sucks, so I don't do it often) them. My primary concern
with any kernel is stability above all else...the 2.4.6-pre1-xfs (with
one hand edited patch to block_dev.c) has been very stable, so I don't
put it on my high priority list to try the latest and greatest.
D. Stimits, stimits@xxxxxxxxxx
|