[lug] minix vs. linux
siegfried at heintze.com
Wed Apr 19 16:22:40 MDT 2006
>>Until Apple OS-X came along, micro-kernels seemed to be used only for
>>things such as real-time applications and embedded systems, and they
>As far as I know, Windows NT used a microkernel design based (though I
>believe it was fairly loosly) off of Mach. This is the same microkernel
>that OSX is based off of. Not that I follow Windows very closely, but
>aren't the current Windows releases based off of Windows New Technology
>Technology, therefore microkernel based? I don't honestly know though...
Can someone explain the merits of the microkernel architecture? I believe it
is supposed to be more stable because system services are implemented in
seperatate processes which can be restarted?
I remember reading about it in a minix book years ago and the author of the
book said that while this business of process context switching for all
system calls (in addition to context switching to inner CPU rings) was slow,
this would be no big deal as processes got faster....
I think I read this back in 1984.
Hmmmph... As far as I can see, the microkernel architecture is crazy.
Asside from all the expensive process context switching windows has to do
(that linux does not) windows has to create a shadow thread for every thread
an application creates. That's nuts! This is so multiple threads in the
system procesess can service multiple threads in the application
And the benefit: Windows is more stable! I guess that is why Windows 3.0 was
not stable: it did not use the micro kernel architecture!
So in case you did not know it: this explains why windows is so much more
stable than linux ;)
More information about the LUG