[Gllug] Multiple X servers
Alain Williams
addw at phcomp.co.uk
Mon Dec 30 11:06:48 UTC 2002
On Mon, Dec 30, 2002 at 09:50:19AM +0000, John Hearns wrote:
> On Mon, 2002-12-30 at 09:26, Alain Williams wrote:
> >
> >
> > That is exactly why I bought a couple of 'spare' video cards a couple of
> > weeks ago. Most PCs spend most of their time doing zilch, and word
> > processing, reading mail, ... are not CPU hogs.
> Agreed.
> > So invest in a reasonable
> > amount of RAM (not much of an investment these days) so that the thing
> > won't swap & I think that you should be able to run several users off
> > one box.
> Agreed - decent amount of RAM a good thing.
>
> But may I (politely) be allowed to think a bit further?
> Your idea is excellent - and fun.
> How about building the box and bringing it along to GLLUG?
> We will all learn something!
OK, good idea - don't hold your breath, I'm busy @ the moment, month or two.
> But for a real classroom etc. you are building in one point of failure.
> Go with the economics you have - but think of thin clients?
That is another way of doing it, but you still have to pay the cost of
the X term. There is no reason why what I am talking about could not
be used as a thin client. The trouble with thin client is that you need beefy
servers & a good/fast network.
I think that, in any case, to make it work/scale well the most that the shared
desktop PC should do is to run the clients (OpenOffice, browser, ...). The user's
$HOME, should be on a central server. Thus if a box does fail (taking several users
with it), a new one is dropped in & they can be back working in 1/2 hour.
> You have only one central box to administer.
> But if there are hardware failures of the clients then (hopefully) you
> have another spare ready to swing in.
Quite. The economics of PCs are that if you have many boxes (say > 10), it is not worth
paying expensive maintenance contracts with things like 4 hour response, ...,
but just have a spare box (they should all be similar spec) so that the faulty
one can be repaired at leisure.
> Please open the debate - I know that the thin client model has failed
> time and time again and I've always wondered really why.
Part of it is I suspect that some users see it as not being a 'real' PC.
I suspect that the multi head/keyboard model that I propose will suffer from
the ''I want to have *my* PC - aren't I important/valued enough for you to
give me *mine* ?'' histrionics from some users.
> However, my argument is that support costs matter too, and thin client
> network.
Agreed. The support costs for a small number of these will be high. If we
can do it a few times, provide reliable/easy HOWTOs, and get the major distros
to support it - costs would drop. -- this is my eventual goal.
Support from the distros would also make it acceptable to the pointy haired types
> Oops, I suppose I ought to be honest and add in hub (£100) and cabling.
> Assume the clients have NICs on board.
I will probably also need to add some extra gubbins in a PCI slot to plug
the extra keyboards/mice into [**]. But I still stand by my ballpark figures.
[**] suggestions on appropriate/cheap hardware please.
--
Alain Williams
#include <std_disclaimer.h>
--
Gllug mailing list - Gllug at linux.co.uk
http://list.ftech.net/mailman/listinfo/gllug
More information about the GLLUG
mailing list