**** BEGIN LOGGING AT Sat May 28 23:59:56 2005
May 29 06:21:52 <[g2]> dyoung-zzzz, from a time perspecitve getting a getting a program loaded and running would be the fastest
May 29 06:22:10 <[g2]> and loading the icache and going would be the quickest route to something like that
May 29 06:22:31 <dyoung-zzzz> Uh huh.
May 29 06:23:00 <dyoung-zzzz> If I knew about these private JTAG instructions, things would be different.
May 29 06:23:06 <dyoung-zzzz> it shows how much I've read the manual!
May 29 06:23:18 <[g2]> so little time so many things to learn about :)
May 29 06:23:56 <[g2]> I'm happy to be absorbing as much as I am
May 29 06:24:14 * [g2] will try harder, but I'm so happy to be plugged in with you guys
May 29 06:24:49 <dyoung-zzzz> I dont really know anything, I just pretend to know what I'm talking about. :-)
May 29 06:26:29 <[g2]> hey you don't know about any pci-express gigE cards do you ?
May 29 06:27:43 <dyoung-zzzz> nope.
May 29 06:29:43 <prpplague> hehe, everyday is a day at school 
May 29 06:29:54 <prpplague> something learn with every project
May 29 06:30:24 <[g2]> prpplague, amen ! :)
May 29 06:31:16 <[g2]> speaking of learning.... I've been playing around with my server box and I've learned quite a lot about its performance
May 29 06:33:19 <dyoung-zzzz> me too.
May 29 06:33:35 <dyoung-zzzz> I've learned that my laptop spanks the hell out of my "fast" desktop.
May 29 06:33:37 <[g2]> dyoung-zzzz, which server are you playing with ?
May 29 06:34:03 <[g2]> I'd love to compare notes about boxes :)
May 29 06:34:09 <dyoung-zzzz> I only have a 3060Mhz P4.
May 29 06:34:22 <dyoung-zzzz> the GigE maxes out at half speed.
May 29 06:34:37 <[g2]> Do you know why ?
May 29 06:34:39 <dyoung-zzzz> so either my clients suck, or its slow.
May 29 06:34:43 <dyoung-zzzz> I didnt look into it yet.
May 29 06:34:57 <[g2]> want the cliff notes ? :)
May 29 06:35:04 <dyoung-zzzz> I was gonna do soda A / soda B with the laptop.
May 29 06:35:21 <dyoung-zzzz> Sure, if you have notes that may be helpful!
May 29 06:35:49 <dyoung-zzzz> I guess what I should do is plug it into itself to see what happens.
May 29 06:35:58 <dyoung-zzzz> (it has 2 GigE interfaces)
May 29 06:36:03 <[g2]> My server setup is an AMD64 1.8G (3000+) 939 on and NForce4 ultra board with 512MB
May 29 06:36:35 <[g2]> It's got on-board gigE and I stuck in both a 3com board ($29) and an EdiMax($14)
May 29 06:36:49 <[g2]> I'm running knoppix 3.8.2
May 29 06:37:20 <[g2]> I can get 600Mbs through the pipe to my P4 with useless data
May 29 06:37:47 <[g2]> I can get nearly 400Mbs of true data across the pipe
May 29 06:38:01 <[g2]> I beleive I'm limited by the PCI bus
May 29 06:38:08 <dyoung-zzzz> I see...
May 29 06:38:32 <[g2]> This is verified by doing the ttcp with a dd if=/dev/zero 
May 29 06:39:13 <[g2]> I can send same about of data (nearly 1 GB) through the stack on the same machine at 2300Mbs
May 29 06:39:53 <[g2]> So the 400Mbs is limited to my disk access which is 60MBs for my single drive
May 29 06:40:34 <[g2]> sending the zeros from /dev/zero I know I'm bumping up against a limit somewhere else
May 29 06:40:55 <[g2]> it's not interrupts because 1500 byte frames or 9000 byte frames are the same
May 29 06:41:13 <[g2]> same for on-board chip and 3com chip
May 29 06:41:32 <[g2]> only the 3com board does the jumbos
May 29 06:42:05 <[g2]> I'm guessing if I striped the drives I could run real data up to 600Mbs
May 29 06:42:52 <[g2]> however, the 400Mbs I'm getting now with a single drive is plenty fast for me as I've got essentially disk speeds from the network
May 29 06:43:14 <[g2]> the processor is only about 20% loaded during this time so I know I've got a bunch of headroom
May 29 06:45:06 <[g2]> Does that help at all ?
May 29 06:52:24 <dyoung-zzzz> It gives me things to look for.
May 29 06:52:47 <dyoung-zzzz> I was thinking my limit was in the actual disk access.
May 29 06:53:09 <dyoung-zzzz> because I have another system on a 12-drive raid that delivers at almost line speed.
May 29 06:53:38 * dyoung-zzzz zzzz's
May 29 06:56:13 <[g2]> dyoung-zzzz, nite.  You can test the theory by sending small files that fit in memory :)
May 29 06:56:48 <[g2]> or by sending something like dd if=/dev/zero then you don't have anything to worry about for producing the data
**** ENDING LOGGING AT Sun May 29 23:59:56 2005