Before you think of being a do-gooder...

[Christopher Browne]

On 5/29/06, James Knott <james.knott-bJEeYj9oJeDQT0dZR+AlfA at public.gmane.org> wrote:
> Evan Leibovitch wrote:
> > Contrast this to the tech fashion inflicted by vendors, such as the
> > current geekism-du-jour, virtualization. This is a creation of vendors
> > to address a problem that doesn't exist for most people -- yet
> > Microsoft, Novell and Red Hat seem to agree that we need it. Most
> > businesses can't keep *one*  OS under control, now vendors are telling
> > them they need to run multiple environments PLUS the supervisor system
> > that runs it all.
>
> One point that IBM made, at the recent Real World Linux conference was
> the idea of running multiple instances of Linux on a large system.  That
> method delivers a lot of economies to someone running multiple servers,
> compared to a bunch of x86 boxes. An example is a virtual lan between
> virtual servers, that runs at memory speed, rather than "only" 1 or 10 Gb/s.

The trouble is that there are some not inconsiderable diseconomies of
scale, as well.

In order to host that bunch of hosts on one physical box, you need to
have a physical box that has a Lot of Gigabytes of Memory, and those
machines are *way* more expensive than beige boxes.

The only way there's an "economy of scale" here is if you can treat a
$100K server as being a sunk cost that, in effect, comes as no cost.

At one point, some of us at work were thinking it might be a sweet
idea to take one of the Opteron systems we have that is underutilized,
and treat it as the "virtual server" for our login sessions and X
clients and such.  But it's not an "economy of scale;" it's an
"economy of using something that would otherwise have been a VERY
EXPENSIVE chunk of wasted hardware."

There are some pretty cool things that could be done with Xen
vis-a-vis process migration, whether for web servers or for database
servers; it is definitely anything but cheap.  And what it *wouldn't*
buy, even at high price, is improved reliability.

You can migrate a database server from one Xen physical host to
another, *if you can plan the migration*.  If the 'former' physical
server falls over, there's no way to grab the last state and
automagically shift it over.   You needed to migrate it *before* the
box fell down.

The result is frankly quite uninteresting.  The case where I most want
migration is the case where it won't migrate.
-- 
http://www3.sympatico.ca/cbbrowne/linux.html
Oddly enough, this is completely standard behaviour for shells. This
is a roundabout way of saying `don't use combined chains of `&&'s and
`||'s unless you think Gödel's theorem is for sissies'.
--
The Toronto Linux Users Group.      Meetings: http://tlug.ss.org
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://tlug.ss.org/subscribe.shtml