RE: [RPRWG] Frame reordering for medium priority traffic in Ganda lf?
Spencer,
Fast retx/rec kicks in after seeing 3 out-of-sequence packet. Now, what
controls do we have at the MAC that limits the out-of-orderness for each TCP
session - for that you will need to maintain per-session context.
One simple example of the effect of reordering and TCP tput drop is as
follow: Suppose a flow has CIR and EIR attributes. The CIR pkts will be
in-profile and EIR portion will be out-of-profile. By causing TCP windows to
fluctuate, what you are doing is impacting the tput of the entire flow -
thus, causing drop in tput in the otherwise "in-profile" packets - thus
maybe consistently violating the svc guarantees for the flow.
BTW, if you are interested, there is a paper by Partridge and Bennett in
IEEE Trans Networking published sometime in 1999/2000 that discusses
reordering in the internet and its effects.
Regards,
Krishna
-----Original Message-----
From: Dawkins, Spencer [mailto:Spencer.DAWKINS@xxxxxxxxxxxxxxx]
Sent: Friday, December 07, 2001 11:14 AM
To: stds-802-17@xxxxxxxx
Subject: RE: [RPRWG] Frame reordering for medium priority traffic in Ganda
lf?
Sorry, but I'm still confused. How can reordering out-of-profile TCP traffic
make things worse? Either it triggers fast retransmit/fast recovery, so the
sending station slows down its sending rate, or it doesn't trigger fast
retransmit/fast recovery, so there is no effect.
Spencer
-----Original Message-----
From: Anoop Ghanwani [mailto:anoop@xxxxxxxxxxxxxx]
Sent: Friday, December 07, 2001 10:05 AM
To: 'Dawkins, Spencer'; stds-802-17@xxxxxxxx
Subject: RE: [RPRWG] Frame reordering for medium priority traffic in Ganda
lf?
First, this is not a single station that we're talking about here.
Second, the out-of-profile traffic was admitted because the
network thought it could deliver it. Reordering the traffic might
actually make things worse. It's like getting a freebie that might
make your life worse without telling you about it.
-Anoop