[e2e] pointers to pitfalls of using FTP to test network performance?

[Amy Wang]

Thank you guys for the reply! 
 
 Yes, I am asking about concrete data in supporting the claimed pitfalls in trying to use application performance to measure network performance. 
 
 For example, how much variability introduced into FTP performance due to disk I/O access, and therefore it is not suitable to use FTP to measure network performance (link capacity). 
 
 I have seen a lot of statements like that for ttcp or iperf related website but I haven't seen concrete data.
 
 thanks,
 Amy


Joe Touch <touch at ISI.EDU> wrote: Amy,

Amy Wang wrote:
> I am trying to collect a list of issues using FTP or windows file 
> transfer to measure network performance in local area network setting. I 
> would appreciate if anyone can guide me to the proper papers and 
> internet links,

If you're trying to measure packet performance, you might try a packet 
generation tool, like netperf or iperf:
http://staff.science.uva.nl/~jblom/gigaport/tools/test_tools.html

If you are trying to measure the performance of transferring a file, FTP 
is a reasonable choice. Testing from a Windows system is useful when 
that's the application you're trying to measure, but you'll probably get 
better results from FTP - even to Windows hosts.

However, I think you're asking more about the pitfalls in trying to use 
application performance to measure network performance, which basically 
translates into "how do I make my apps use the network well" - for that 
purpose, Matt Mathis' work at PSC would be useful to consult:
http://www.psc.edu/networking/projects/tcptune/

Joe


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.postel.org/pipermail/end2end-interest/attachments/20080228/ef80c155/attachment.html