[ltp] Slightly OT: wifi througput, overhead and goodput?

[Jiang Qian]

On Sat, Mar 17, 2007 at 08:32:23PM +0100, Stefan Blum wrote:
> Hi, though I am not at all a network guru, I recently read an article in a 
> German IT-newsmag. Here's a short courtesy translation of the paragraph 
> relating to your question:
> 
> "The transmission of a payload of 1500 bytes with 54 MBit/s takes 325 
> microseconds, equaling a net-rate of approx. 37 MBit/s. If you take into 
> account TCP/IP-Overhead (additional 40 bytes per packet, TCP-ACKs) and 
> repeated transmissions due to radio-interferences, you get net-rates of 
> approx. 25 MBit/s, which 802.11a usually acchieves on a good 
> radio-connection. ... For 802.11g, that in principle uses the same radio 
> technique as 11a, it might get worse."
> (German original: http://www.heise.de/netze/artikel/80521/0)
> 
> It seems like you get pretty much the max out of your 54MBit/s connection if 
> you have a throughput of about 24MBit/s.
> 
> Hope that helps! Regards,
> 
> Stefan
Thanks a lot Stefan.
Jiang
> 
> 
> On Friday 16 March 2007 04:09, Jiang Qian wrote:
> > Hi all:
> > 	To all the network gurus out there, these questions has been
> > bugging me for a while and haven't been resolved through all the google
> > search and wiki reading. I've been trying to make a sense of how good
> > the ipw2200 card of my T43 talk with my WRT54GL router and what is the
> > possible bottleneck of the network.
> >
> > 1. I read on wikipedia[1] that the maximum throughput of 802.11g is
> > 24.7Mb/s, in contrast with the maximum data rate of 54Mb/s. How does the
> > discrepancy come about? I thought this has something to do with only
> > half of the channel is available, but that should give 27Mb/s.
> >
> > 2. As I read here[2], even the throughput is not the unidirectional data
> > transfer rate. Goodput is a better measure. For example, due to tcp
> > overhead, only 94.92Mb/s is possible for the 100Mb network. So does the
> > similar overhead apply to wireless network? Or does the wireless network
> > have an even bigger overhead? Is this overhead already taken into
> > account in the reduction from 54 to 24.7Mb/s in question 1.?
> >
> > 3. To put it all together, when I download one 131MB file via ftp from
> > my desktop(connected to wired port of router) to my laptop(connected to
> > the wireless port), the log of vsftp report a download rate of
> > 2893Kbyte/sec.  That is 2893*1024*8/1,000,000=23.7Mb/sec. This is the
> > one direction effective goodput[2], I suppose. What percentage am I from
> > the theoretical maximum of the unidirectional tcp goodput, combining
> > consideration of the previous two points?
> >
> > 	Any comment or enlightenment would be most appreciated. If you
> > find wikipedia entries cited here to be wrong, please correct them or I
> > can do the typing for you:)
> > 								  Jiang
> >
> > Reference:
> > [1]http://en.wikipedia.org/wiki/802.11g#802.11g
> > [2]http://en.wikipedia.org/wiki/Goodput
> -- 
> The linux-thinkpad mailing list home page is at:
> http://mailman.linux-thinkpad.org/mailman/listinfo/linux-thinkpad