Re: [Etherboot-discuss] Throughput issues with gpxe
Brought to you by:
marty_connor,
stefanhajnoczi
From: Miller, S. <Sha...@yr...> - 2009-05-29 02:58:29
|
Ok. I'm sorry about the redundancy. I only offered that because I thought you would want to get past your hanging error and actually get to try out whatever TCP window size you had chosen. It seems that you aren't even getting the chance to try your choice, since you aren't getting to the point of downloading tftp://${next-server}/pxelinux.cfg/default. Does the information at http://www.ncsa.uiuc.edu/~vwelch/net_perf/tcp_windows.html help at all? In there, it suggests a formula for choosing a TCP window size. I sort of suspect that your original question would have an answer something like: a) gPXE's current value should be functional (lots of folks use it) and b) tuning it to be optimal depends on your situation... I'm not sure there's a good Universal "optimal" beyond what's in place. Sorry I couldn't help. - Shao Miller ________________________________ From: Mario Spinthiras [mailto:spi...@gm...] Sent: Thursday, May 28, 2009 21:17 To: Miller, Shao Cc: eth...@li... Subject: Re: [Etherboot-discuss] Throughput issues with gpxe Hello Shao, While I would like to thank you for a reply I would like to inform you that the information you provided is clear from RTFM from the website. There is nothing new to that. My question still stands clear in its context ; Has anyone played with the hardcoded value of the tcp window size in tcp.h of the gpxe sourcecode? Regards, Mario. |