[K12OSN] what do these iptraf results tell me

[Carl Keil]

Hi Folks,

I'm running K12LTSP 5 on an MSI motherboard, Dual Core CPU, PATA drives, 
one for /home one for everything else.  I have eth0 plugged into an 
unmanaged, consumer grade gigE switch.  I have several different thin 
clients plugged into the switch. 

I asked the list previously for suggestions about how to check what 
actual network performance I'm getting.  Someone suggested I run TuxType 
fullscreen while running IPTraf. 

When I run IPTraf on eth0 I see around 50Mb/s of traffic with one client 
running TuxType.  I've gotten two clients gotten 2 clients going playing 
battle for Wesnoth and gotten the traffic on the server up to about 
70Mb/s.  So I know the NIC on my server can serve more than 50Mb/s, but 
I thought that TuxType fullscreen took around 70.  At 50 Mb/s, the 
performance in TuxType is pretty terrible.  The screen redraws are 
jerky, the letters don't come down very fast or smoothly.  There's a ton 
of latency when selecting even from the menu to start the game.  What 
does this tell me about where my bottleneck might be?  CPU and RAM are 
way below 50%. (I can't remember the exact numbers) No swapping.  It 
seems like all that's left is disk, but I can't imagine that Tux or 
Wesnoth is that disk intensive in the midst of a game.  Or the wiring to 
the clients. 

Any thoughts about how to track this down?  My clients are kindof 
whimpy, do I need more than 4 megs of video ram?  Does the type of video 
card make a difference?  What's best?  Better nics in the clients?  I'm 
kindof confused.  I thought you could get good performance from dumpster 
hardware using K12LTSP.  By and large I'm very happy with my thin client 
setup, but this is one area I'd like to improve. 

Just to be clear, my main problem is slow screen redraws during 
fullscreen games where the whole screen is moving.  I get the same 
effect when I scroll a web page with the browser window maximized.  It 
just scrolls slowly and choppily.  Any ideas what's the limiting factor 
here?

Thanks,

ck