On Sun, Apr 14, 2002 at 03:15:39PM -0400, Lamar Owen wrote:
> ISTM that improving interactive performance would also improve multiuser
> performance in a server, as low latency and kernel preemption can increase
> multiuser server responsiveness.
I doubt any performance will increase, either on a multiuser or on a
singleuser system.
Having faster response on mouse clicks or keyboard input doesn't translate
on better overall performance, the user just has the felling that it's so.
As an example, a part of those patches causes brakes in the middle of some
loops (saving buffers to disk, etc). Then other applications that don't
depend on disk activity can have change to run, so the system seems
faster, it's more responsive. But it won't actually be faster, the system
still has to lock again and continue saving the buffers. Actually, in this
case there will be an overhead caused by checking if the kernel should
brake.
However, both projects review the Linux code, and may find, if they
haven't already, some places were a finer locking may be used, giving a
better performance in a SMP system. But it could also break some
integrity.
Those patches are not recomended for a server, and now I'm curious to
check if the -enterprise configuration has them active.
> Did you happen to report it to Red Hat's Skipjack list, or to
> bugzilla.redhat.com/bugzilla? Helps make a better dist!
Alas, a bug report saying: the system crashed, I can't login remotely,
doesn't help a lot...
Regards,
Luciano Rocha
--
Luciano Rocha, strange@nsk.yi.org
The trouble with computers is that they do what you tell them, not what
you want. -- D. Cohen
