[e2e] Short Fat Networks, tcp and Policers

=?gb2312?q?Jing=20Shen?= jshen_cad at yahoo.com.cn
Wed May 12 04:35:25 PDT 2004


I've got some experience with providing bandwith by L3
traffic shaping. But, in one of those experients
people complaints the degradtion of web site access
speed after upgrading from 10Mbps ethernet channel to
a 20Mbps one. In that experiments  RTT from that site
to primary game site falls in (10ms, 50ms), and  20%
of traffic is UDP. 

We haven't got enough time to look into details of the
problem,  but to my understanding this is due to
packet dropping at policing point and the interaction
between bursty UDP traffic  & TCP flows. 

Is that possible to expose more information on your


Jing Shen

> F) What is the point?!
> ======================
> The points are 
> 1- policers are much easier to build that shapers,
> so we should start to understand them.
> 2- acceptance tests are usually done with very short
> RTT times, so that the timeout phase is
> quite relevant.
> 3- All these exponential smoothings and estimations
> of RTT RMS are useless if the smallest timeout
> is 300msec, and typical LAN's are less that 50msec.
> 4- alot of TCP improvements have been based around
> "LFN"'s but it may turn out that alot of the broad
> band networks are really "SFN"'s (Short Fat
> Networks). 
> G) The End
> ============
> Thank you for you patience, comments are welcome.
> Yet Another Simpleton

Do You Yahoo!? 

More information about the end2end-interest mailing list