4.2bsd kernel auto-nicing, scheduling

[Michael Wagner]

In article <172 at epimass.UUCP> jbuck at epimass.UUCP (Joe Buck) writes:
>I don't believe it!  I, a mere mortal, get to correct Guy Harris :-).

I don't think so.  At least not this time.

>Though you frequently see the word "algorithm" associated with schedulers,
>they are not because they don't terminate (at least they aren't supposed
>to).  

I certainly think schedulers terminate.  There might be some confusion about
either what a scheduler is or what time scale to examine their behaviour
on.  Schedulers that I am familiar with are entered to decide who to run
next, and return (i.e. terminate) with a recommendation.  If it were 
otherwise, you'd never get any useful work done.  The scheduler does,
usually, keep some sort of historical data around to assist in it's 
decisions, but that doesn't change the fact that it doesn't run all the
time.

>
>Of course, the function that decides the next process to be executed at
>a given time is (I hope!) an algorithm.
>-- 
>- Joe Buck <ihnp4!pesnta!epimass!jbuck>


The function that decides the next process to be executed at a given time
is (I hope) a scheduler!

Michael Wagner (wagner at utcs)