Re: Comments on the KSE option

At 12:06 PM +0800 10/28/06, David Xu wrote:
On Saturday 28 October 2006 11:51, Julian Elischer wrote:
> Alexandre "Sunny" Kovalenko wrote:
> >
> > I apologize for misinterpreting your words. But then, if I have
> > M:N set to 10:1, I would expect application with 1000 process
> > scope threads to have as many CPU slots as 100 processes, or,
> > if I have 10 system scope threads and 990 process scope threads,
> > I would expect application to have as many CPU slots as 109
> > processes. Is this what you refer to as "fairness"?
>
> M:N is not a ratio, but rather the notation to say that M user
> threads are enacted using N kernel schedulable entities (kernel
> threads). Usually N is limited to something like NCPU kernel
> schedulable entities running at a time. (not including sleeping
> threads waiting for IO)
> (NCPU is the number of CPUs).
>
> so in fact M:N is usually M user threads over over some number
> like 4 or 8 kernel threads (depending on #cpus) plus the number
> of threads waiting for IO.
>
Julian

As you are emphasizing fairness, I must say I don't believe fairness
in libpthread either, I don't think writting a fairness scheduler is
an easy work, does kernel have made fairness for threads in same
ksegrp, so does libpthread's userland scheduler ? they don't, it can
make threads in same ksegrp misbehaviored, so what we have done is
still process scheduling fairness even there is ksegrp in kernel,
and now sacrificed fairness between threads.

I think we have different ideas of what is the goal is with this
claim of "fairness".

If I understand it right, it is *not* that some static code in the
kernel is going to decide which applications are fair and which ones
are not fair. IIIRC, what Julian wants to do is provide a way that
the administrator can make that decision. The administrator will
have a way to throttle some thread-crazy process, but only if the
*administrator* wants to do that.

If the machine is for a single user, then that user will probably
set "N" to a high value. But if the machine has a lot of users on
it, then the administrator of that machine may want to set "N" to
the number of CPU's on the system, or maybe the number of CPU's
minus 1. And if the users don't like that, then they can go buy
their own damn machine instead of using the machine someone else
bought and is allowing them to access for free.

At RPI we have both kinds of machines. Machines owned by a single
user, and machines which have multiple students ssh'ed into at the
same time. I can see wanting to throttle thread-crazy processes on
some machines, and not wanting any throttling at all on others.

...but it has been a few years since the presentation that I remember
Julian giving about this, so maybe I am not remembering it correctly.
I do remember that whatever it was, it sounded pretty reasonable at
the time. :-)

--
Garance Alistair Drosehn = gad@xxxxxxxxxxxxxxxxxxxx
Senior Systems Programmer or gad@xxxxxxxxxxx
Rensselaer Polytechnic Institute or drosih@xxxxxxx
_______________________________________________
freebsd-current@xxxxxxxxxxx mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "freebsd-current-unsubscribe@xxxxxxxxxxx"

Relevant Pages

  • Re: Comments on the KSE option
    ... A naive scheduler would give ... limits in place for system scope threads. ... what you are referring to above), and "fairness" (not giving process ... are enacted using N kernel schedulable entities. ...
    (freebsd-current)
  • Re: call for more SD versus CFS comparisons (was: Re: [ck] Mainline plans)
    ... problems on my Amarok machine. ... usually test Amarok playback while kernel package make, ... Putting load onto the VM layer / block layer. ... also the behavior on a server for example - how does the scheduler behave ...
    (Linux-Kernel)
  • Re: page fault in sched_pin()
    ... > sched_pinearly in the boot process. ... I wasn't booting off the kernel I thught I was booting off. ... remove it from the scheduler inteface and make it part of the standard ... page fault while in kernel mode ...
    (freebsd-current)
  • Re: BIND9 performance issues with SMP
    ... instead of having its own dedicated scheduler activation. ... that threads that block in the kernel don't block the whole process. ... but lock contentions seem particularly heavy ... The first-level optimization is to create ...
    (freebsd-current)
  • Re: [OT] Interview with Con Kolivas on Linux failures
    ... Con Kolivas, the kernel hacker who authored a better scheduler, recently decided to quit. ... Loss for Linux ... The former option means you have a CPU scheduler which is difficult to model, and the behaviour is right 95% of the time and ebbs and flows in its metering out of CPU and latency. ...
    (Debian-User)