From: WorkerBee (ciscobee@gmail.com)
Date: Sun Sep 30 2007 - 23:53:59 ART
UDP does not negotiate TCP Tranmission window size. So dropping UDP
based on congestion output queue is not related to RED/WRED.
Since no 2-way setup is used by UDP, UDP is best suitable for lossy
environment and the advantage is, less overhead and more efficient
protocol under lossless environment.
On 10/1/07, Michael Brooks <michaeljbrooks@gmail.com> wrote:
> Thanks Ryan. I have seen that document as well, but I am suprised a
> document is not published specifially stating that "RED will drop UDP
> packets" especially when UDP is on every network.
>
> Regards,
> Mike Brooks
> CCIE# 16027
>
>
> On 9/30/07, Ryan Morris <ryan@egate.net> wrote:
> >
> > It drops packets based on queue depth. It doesn't care what kind of IP
> > packet it is.
> >
> > Although it doesn't say it explicitly, the answer is here:
> >
> >
> > http://www.cisco.com/univercd/cc/td/doc/product/software/ios122/122cgcr/fqos_c/fqcprt3/qcfconav.htm#wp1000893
> >
> > "In considering the usefulness of RED when robust transports such as TCP
> > are pervasive, it is important to consider also the seriously negative
> > implications of employing RED when a significant percentage of the traffic
> > is not robust in response to packet loss. Neither Novell NetWare nor
> > AppleTalk is appropriately robust in response to packet loss, therefore
> > you should not use RED for them."
> >
> > UDP is a non-robust transport, and would not benefit (in fact could suffer
> > greatly) from having a large number of random packets dropped.
> >
> > Regards,
> >
> > Ryan Morris
> > CCIE #18953
> >
> > On Sun, 30 Sep 2007, Michael Brooks wrote:
> >
> > > All,
> > >
> > > I have a $1 dollar bet with one of my colleagues that RED or even WRED
> > > (depending on how its configured) does drop UDP packets. He states that
> > > since the congestion avoidance mechanisms of RED only works on TCP flows
> > > that only TCP packets will be dropped "only". The problem is I have
> > been
> > > searching and can't find any good documentation to back up my
> > theory. Also,
> > > I haven't had time to test in the lab. Does anyone have some good
> > > links proving one of our theories (preferably mine ;-) ). Your input
> > would
> > > be greatly appreciated.
> > >
> > > Thank you,
> > > Mike
> > >
> > > _______________________________________________________________________
> > > Subscription information may be found at:
> > > http://www.groupstudy.com/list/CCIELab.html
> >
> > _______________________________________________________________________
> > Subscription information may be found at:
> > http://www.groupstudy.com/list/CCIELab.html
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
This archive was generated by hypermail 2.1.4 : Fri Nov 16 2007 - 13:11:11 ART