[Info-vax] OS design, was: Re: [OT] Wirth style languages, was: Re: Obscure Ada compiler vendors?
Simon Clubley
clubley at remove_me.eisner.decus.org-Earth.UFP
Tue Apr 9 13:35:54 EDT 2013
On 2013-04-09, David Froble <davef at tsoft-inc.com> wrote:
>
> Of more concern to me is VMS (as far as I know) is more of a monolith,
> with lots of complexity and inter-relationships. Is this reality?
Yes. There are inter-relationships in VMS which simply do not exist in,
say, Linux, even though both are monolithic kernel designs.
> Any
> changes can have far reaching ramafications. Is this the best way to
> put an OS together?
In the 1970s, yes. In the 21st century, no way.
Even Linux, which is a monolithic kernel, not a microkernel, has clean
separation between the generic and architecture specific parts of the
kernel as well as the other parts of Linux. (Compare adding a filesystem
on VMS with adding yet another filesystem into Linux, for example.)
Linux also makes minimal demands on the capabilities of the host hardware
when compared to VMS which demands much more functionality from the
hardware it runs on.
If you want proof of those statements, just look at the large range of
architectures which Linux runs on. Now compare that to the situation
with VMS and the feedback that porting VMS to a new architecture is a
enormous exercise.
And no, I am _NOT_ spreading doom and gloom; I am simply trying to wake
people up and make them see what is out there and what a portable VMS
will have to compete with. If you come up with a way to still make that
happen then good for you; having more viable OS options can only be a
good thing.
> I sure don't know. But I read about micro-kernels,
> and that path has advocates.
>
> I'd think that a more modular approach might avoid some issues that now
> might be a problem in VMS. Don't really know. I also don't know if a
> more modular approach might be very detrimental in some ways, such as
> performance.
A more modular approach in a monolithic design (ie: Linux style) should
not see any major dropoff in performance if done correctly.
Microkernels do have performance impacts, but all performance is relative.
Furthermore, you can have a microkernel which still has hard realtime
capabilities overall. For example, QNX is a microkernel based design; see
http://en.wikipedia.org/wiki/QNX for further information on QNX.
Simon.
--
Simon Clubley, clubley at remove_me.eisner.decus.org-Earth.UFP
Microsoft: Bringing you 1980s technology to a 21st century world
More information about the Info-vax
mailing list