[nzlug] expanded workstations vs thin clients

[Simon Bridge]

On Tue, 2008-01-22 at 09:01 +1300, Cliff Pratt wrote:
> Simon Bridge wrote:
> >> There's no saving in space, energy or hardware with multiple keyboards,
> >> monitors and mice.
> > 
> > Um... compared with, say, running 10 tower desktops as well as the
> > monitors etc???
> >
> You need a beefy machine, plus it will run multiple video cards. It also 
> appears to use single run VGA and USB to run the workstations. That 
> severely limits where the master and the workstations are situated (ie 
> close to the master).
> 
Well... one point at a time:
Beefy machine: is this going to power or space or both... I can run 8
workstations off the machine I have on my desk... so, for the sake of
concrete comparison, the systems would be 8PCs vs 1PC+7thin vs 1PC+4VGA
cards.
... Space, using 8 PCs would fill the available desk. Thin clients can
be much smaller of course, though they still take up some room - the
"space saving" is much less. (ECE has dell thin clients strapped behind
tft monitors - effectively half the space of a typical CRT between them,
and the desktop footprint is little larger than the monitor stand.)
... Power: granted the video-card tends to be the big power consumer.
Presumably thin clients have video-cards too? So there's no expected
a-priori saving on those. 8PCs have a bunch of extra stuff to run... the
expanded solution is a clear advantage here. Thin clients also need
power supplies... so adding 7 of these presumably increases power
consumption over not having them (and putting the VGA cards in the
central PC.)

The exact saving in power and space will depend on the exact hardware
employed and the relative workloads. It can be argued that the savings
in favour of expanded systems are small, fine, but this is different
from claiming that they are non-existent.

There will exist situations where such savings are small compared to
other comparative features. However, there will also exist situations
where it is the other way around.

> It would be no good in the family situation where the 'work stations' 
> are unlikely to be located close together. (By 'work station' I mean the 
> place that the person works, not a machine)
This is a different argument: I was taking issue with the statement that
there is no saving in space or power to be had from the expanded
workstations.

You meant: no saving over xterminal boxes, OK, but the comment you
responded to was not comparing with them, but with a 1PC/seat solution.
So I take issue with that statement on two fronts: (a) It misrepresents
the intent of the original statement and (b) it is false: there *are*
space and power savings to be had over thin solutions.

> > 
> It still seems to me to be useful only in restricted situations.

The same can be said for *all* IT solutions, and nobody is claiming
otherwise. Even a niche solution has value.
Clearly laying so much cable through a household or corporate offices is
not practical. This just means that power and space are not the only
thing to be considered here.

The expanded solution become practical where stations would normally be
close together, like a school, a lab, an internet cafe, or a kiosk set. 

Especially suited are situations where there is already a substantial
investment in fat-client solutions and the aim is to expand and/or
"upgrade" the system. Also where there is no existing investment in
ethernet. (Extra-long cables are not essential either: if you put the
stations in a circle around the PC, all the cables will reach.)

These deployments embody the vast majority of the ways that general
public interact with computers if not exactly a significant proportion
of all deployments.

There are home deployments (I've seen them) where two or more computers
share a table... it would suit here too. So, there exist, today, some
"family situation"s where this solution would be practical.

Similarly, how good is the solution you have in mind for playing online
games? Families with at least one gamer, and/or who stream a lot of
video, will find thin-client deployment impractical if it cannot keep up
with (so called) rich content.

The "family situation" is not as clear cut as some people would have it.
It tends to favour single fat clients (1PC/station) for reasons other
than marketing (marketing too).  While the expanded station is not often
useful, the xterminal is not either. This can be used to support the
argument that one is no advantage over the other, but only in the (weak)
sense that they are both bad.

You may think this is over-pedantic, after all, *I* am sometimes
imprecise. But consider the statement you responded to...

................
The articles I linked in the previous post have some problems. For eg.
for some reason the author thinks you need twice as many PCs as seats
for the fat solution. The table does not include additional (long) vga
cables, nor indication of how the mouse and kbd get connected.

It does say that you need special software to run this, but I don't see
why you do.

But - hey, it's advertising really.

Distributed maintenance of thin clients is not really a problem, as you
need distributed maintenance of HIDs anyway. With a thin solution there
is less cabling, which would argue for lower, long-term, maintenance
costs.

There are many considerations in a cost-benefit analysis than were
considered in the article.