Sheldon Hearn wrote (2001/02/01):
> What I remember of the discussions that surrounded this one, your
> summary is correct. The only thing is that nice isn't so much _broken_
> as it just isn't doing what you'd expect it to. :-)
Ok, scheduler in -current is not broken. But I'm afraid that in -stable
it is - can (niced) process cause a lock of machine?. Currently, we
have dual processor box with 4.2-STABLE and it silently locks too often.
With current scheduler in mind, it is hard to say if I should search in
HW or SW for a potential fix...
> I don't think any of the FreeBSD manual pages suggest that nice 20
> processes aren't supposed to get _any_ CPU time.
Maybe. But there are some conventions and two-process sensitivity
2.5:1 is not very big (is low). Solaris and Linux have much bigger
ratio (sufficient). So why FreeBSD has to be so much different?
Insensitivity of nice were problem in the past and it is going back.
Users need to run some processes (!= dnetc != setiathome) enough hidden
and idprio does not work either. Unfortunately, it is sufficient
consideration why do not use FreeBSD for example as a general purpose
server: When anybody needs to run long-time computation process, he
can use only nice 20 and other users are not blind and can see, that
operating system does not want to give sufficient time to their
short-time computation processes. And users can force to switch
to another system - just hear "but Linux do..." :-(
Rudolf Cejka ([EMAIL PROTECTED]; http://www.fee.vutbr.cz/~cejkar)
Brno University of Technology, Faculty of El. Engineering and Comp. Science
Bozetechova 2, 612 66 Brno, Czech Republic
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-current" in the body of the message