On Sun, 27 Apr 1997, David Williams wrote:
> In message <Pine.HPP.3.94.970426174342.20756A-100000@emmy.smith.edu>,
> Michael Callahan <mjc@stelias.com> writes
> >
> >I agree with what Alan, David and others have said about 1000s of threads
> >being the wrong answer to the question--essentially no matter what the
> >question is. (This is assuming you don't have a machine with 1000s of
> >CPUs!)
> >
> Exactly - If you only have 5 CPUs you can only ever have 5 threads
> actively run on CPUs at the same time. Perhaps a few more to handle
> disk I/O, one if you use Kernel Asyncchronous I/O and a few to handle
> network actively, Informix RBBMS recommond one per 200-300 users i.e.
> clients.
>
> Why have more??
>
Well, why not extrapolate that idea even more? You can only have one
process per processor, so what's the point of multi-processing (beyond the
number of physical processors you have). Why not use a sophisticated
event scheduler, and ... wait, someone thought like this long ago on Intel
platforms, and created a not-so-attractive GUI based on their thoughts.
> --
> David Williams
>
+-------------------------------------------------------------------+
+ -- Finger: flood@evcom.net for my PGP public key -- +
+-------------------------------------------------------------------+
-----BEGIN PGP SIGNATURE-----
Version: 2.6.2
iQCVAwUBM2SmQhsjWkWelde9AQFyMgP/aJGibwB2PHMc15onQlF5B35Fyz8F98Hi
iFqjrD6yXHmGLLjJ7kRijdtam+JAT8hx7BRvX3Jiqx5RmmaOnNIAZAr0CmiqyZ6O
hj+7+KzftVgtEa0O3C/58VzXd6MXEA1IgnpV072VEAZCjBcxw8Wj0oIHN3fHDvri
U9LA8WpBUXs=
=uHCe
-----END PGP SIGNATURE-----