Virtualization, 20yrs late. Was: "Death of a firewall" and Xen

Andrew Lentvorski bsder at allcaps.org
Mon Jul 11 09:13:27 PDT 2005


Tracy R Reed wrote:
> Yeah, yeah. The rant of the curmudgeon. I am quite familiar with it. :)
> Were they doing it for $300 a box and giving you the source? I don't
> think so. And that is a very different thing which make FOSS "innovative".

No, but unlike FOSS, it is tested for backward compatibility for 30+ 
*years*.

Personally, I'd take an actual unit tested OS over more features at this 
point.

> Yes, it certainly would. Unfortunately the mainframe world has never
> been all that good about cooperating or sharing knowledge. The Unix
> world is having to reinvent the wheel and the Free software world is
> having to invent it again. In spite of all of this, progress is being
> made and we can now have virtualization on our $300 PC's.

Ahem.  The pee-cee world has never been particularly good at adhereing 
to a standard.  In addition, the x86 architecture has been absolute 
garbage for having the extensions required to support some of this stuff 
until the last couple of years.

For the love of Deity, I *still* can't do remote access and control of a 
$300 pee-cee.  IPMI compliant stuff is starting to appear, but it's just 
*starting* and its expensive.

It isn't the fact that mainframes hoarded their knowledge (I can point 
you to *lots* of IBM Journal of R&D articles on how to do 
virtualization).  It's simply that none of the FOSS folks cared until 
recently.

Personally, I *still* don't care.  Rather than virtual x86's, I would 
rather have a "virtual processor" so I don't have to care about x86, 
PowerPC, Alpha, etc.  I'm tired of "processor specific".  Given that 
people don't actually care about overall performance in 90+% of all 
applications, it's time we dumped that whole issue in the trash.

-a



More information about the KPLUG-List mailing list