Discussion:
[Ltsp-discuss] Where does "fat client" begin?
Rolf-Werner Eilert
2015-12-14 08:11:29 UTC
Permalink
Hi folks,

We have a language school and two labs with about 20 places each. We
have been working with an LTSP 4.2 for quite some time now, and it's
time to think about upgrading our terminals to be able to use LTSP 5.
Before I start looking for hardware, I would like to know where you
would draw the line between a thin and a fat client.

On our current setup, I learned that transferring the graphics to the
terminals is a bottleneck. Another problem are programs like browsers
which tend to suck a lot of graphical data into the terminal's RAM
(large pictures for instance), so after some time the terminals start to
swap. You know what I mean... Another problem are terminals with no
modern graphics acceleration - like ours.

We would like to use a current KDE desktop and up-to-date browser like
Firefox, Wine etc. So I thought it might be better to let the terminals
each have their own complete OS booting and use a common pool of
individual and public defaults from the server. Maybe just using the
binaries from the individual harddiscs, but deviating all other
directories to those on the server.

But would I need LTSP for such a thing? Would that still be a fat
client, or: how do you define a fat client under LTSP 5? And would you
think thin clients would do? (Personally, I would prefer thin clients.)

Thanks for any opinion :)

Regards
Rolf
E Kogler
2015-12-14 08:23:53 UTC
Permalink
A Thin-Client has no OS & applications installed, a Fat-Client has.
Von: Rolf-Werner Eilert <eilert-***@t-online.de>
An: ltsp-***@lists.sourceforge.net
Gesendet: 9:11 Montag, 14.Dezember 2015
Betreff: [Ltsp-discuss] Where does "fat client" begin?

Hi folks,

[...]

But would I need LTSP for such a thing? Would that still be a fat
client, or: how do you define a fat client under LTSP 5? And would you
think thin clients would do? (Personally, I would prefer thin clients.)

Thanks for any opinion :)

Regards
Rolf
Rolf-Werner Eilert
2015-12-14 09:07:54 UTC
Permalink
I see, so the definition is clear to me. Thank you!

Rolf
Post by E Kogler
A Thin-Client has no OS & applications installed, a Fat-Client has.
------------------------------------------------------------------------
*Gesendet:* 9:11 Montag, 14.Dezember 2015
*Betreff:* [Ltsp-discuss] Where does "fat client" begin?
Hi folks,
[...]
But would I need LTSP for such a thing? Would that still be a fat
client, or: how do you define a fat client under LTSP 5? And would you
think thin clients would do? (Personally, I would prefer thin clients.)
Thanks for any opinion :)
Regards
Rolf
------------------------------------------------------------------------------
_____________________________________________________________________
https://lists.sourceforge.net/lists/listinfo/ltsp-discuss
For additional LTSP help, try #ltsp channel on irc.freenode.net
------------------------------------------------------------------------------
_____________________________________________________________________
https://lists.sourceforge.net/lists/listinfo/ltsp-discuss
For additional LTSP help, try #ltsp channel on irc.freenode.net
Alkis Georgopoulos
2015-12-14 08:55:28 UTC
Permalink
In the LTSP terminology:
* thin clients are diskless. They load a minimal OS from the server
via NBD/NFS, and at login time they connect to the server with "remote
X" to run the applications on the server, and transfer the screen over
the network.
* fat clients are diskless. They load a full OS from the server via
NBD/NFS, and at login time they connect to the server only for
authentication and to mount /home with SSHFS/NFS. They user their own
CPU/RAM to run the applications.

My advice for you is to buy normal clients, e.g. core i3 with 4 GB RAM
or similar NUCs, but without hard disks, and then follow the ltsp-pnp
installation method:

https://help.ubuntu.com/community/UbuntuLTSP/ltsp-pnp
Post by Rolf-Werner Eilert
Hi folks,
We have a language school and two labs with about 20 places each. We
have been working with an LTSP 4.2 for quite some time now, and it's
time to think about upgrading our terminals to be able to use LTSP 5.
Before I start looking for hardware, I would like to know where you
would draw the line between a thin and a fat client.
On our current setup, I learned that transferring the graphics to the
terminals is a bottleneck. Another problem are programs like browsers
which tend to suck a lot of graphical data into the terminal's RAM
(large pictures for instance), so after some time the terminals start to
swap. You know what I mean... Another problem are terminals with no
modern graphics acceleration - like ours.
We would like to use a current KDE desktop and up-to-date browser like
Firefox, Wine etc. So I thought it might be better to let the terminals
each have their own complete OS booting and use a common pool of
individual and public defaults from the server. Maybe just using the
binaries from the individual harddiscs, but deviating all other
directories to those on the server.
But would I need LTSP for such a thing? Would that still be a fat
client, or: how do you define a fat client under LTSP 5? And would you
think thin clients would do? (Personally, I would prefer thin clients.)
Thanks for any opinion :)
Regards
Rolf
------------------------------------------------------------------------------
_____________________________________________________________________
https://lists.sourceforge.net/lists/listinfo/ltsp-discuss
For additional LTSP help, try #ltsp channel on irc.freenode.net
Helmut Lichtenberg
2015-12-14 09:12:25 UTC
Permalink
Hallo Rolf-Werner,
Post by Rolf-Werner Eilert
On our current setup, I learned that transferring the graphics to the
terminals is a bottleneck. Another problem are programs like browsers
which tend to suck a lot of graphical data into the terminal's RAM
(large pictures for instance), so after some time the terminals start to
swap. You know what I mean... Another problem are terminals with no
modern graphics acceleration - like ours.
We would like to use a current KDE desktop and up-to-date browser like
Firefox, Wine etc. So I thought it might be better to let the terminals
each have their own complete OS booting and use a common pool of
individual and public defaults from the server. Maybe just using the
binaries from the individual harddiscs, but deviating all other
directories to those on the server.
But would I need LTSP for such a thing? Would that still be a fat
client, or: how do you define a fat client under LTSP 5? And would you
think thin clients would do? (Personally, I would prefer thin clients.)
we had lot's of trouble with older thin client hardware (Neo, Levigo, Terra),
especially with bigger monitors and higher resolutions. We also use KDE,
Iceweasel, and everything users might want to have.

After some years I was tired of struggling with these hardware problems and
XWindows (xorg). But I wanted to stick to thin client as I like this concept
and it keeps (well, should keep :^) maintenance low.

So we kept the best pieces of the old hardware and replaced the others with Zotac
Zbox Mini PC (C1320Nano). These are nice little boxes, fanless, quad core,
recent Intel graphics on board.

We run them mostly as thin clients (Debian, LTSP 5.5.4), but we provided a few
of them with an SSD and use them as standalones.

Viele Grüße
Helmut
--
-------------------------------------------------------------------------
Helmut Lichtenberg <***@fli.bund.de> Tel.: 05034/871-128
Institut für Nutztiergenetik (FLI) 31535 Neustadt Germany
-------------------------------------------------------------------------
Loading...