Biaz,
This is very true. For transports, like TCP, that have no way to disambiguate
the receiver's internal load delay from the network's delay, RTT is only
inaccurately and variably measureable. The same thing was said of the
retrans-timeout confusion of cause of loss -- no way to distinguish a
congested router vs errored link. Then we add to this the various ways
transports look at timer ticks under various OSs and even more uncertainty
enters the observations.
Ways to determine true causes include transport agents at the receiver and
intermediate routers. This is something like what has been suggested for
various transport revisions, and system-wide extensions, to measure and
guarantee service more intelligently. Under those circumstances, a modelling
technique, using appropriate filtering of measuremnents, would make sense.
Otherwise, we'd end up filtering what we'd expect to be noise, with certain
properties, yet be ignorant of actual correlations and tendencies in that
'noise' which, in fact, invalidate our filtering from the start.
I also agree with the comments made about proxies being the way networks can
address protocol improvements without overall network deterioration.
Alex
Biaz Saad wrote:
>
> On Wed, 27 Jan 1999, Phil Karn wrote:
>
> > I still think somebody with a better handle on control theory than me
> > (like someone who really knows Kalman filtering) could solve this
> > problem. It's still worth doing because it'll take time to get ECN
> > widely deployed.
>
> Kalman filtering would work if there exists in your signal a true
> "tendency or behavior". Unfortunately, the RTT variations in response to
> the variation of number of packets in flight are most of the time
> uncorrelated. For example, increasing the number of packets in flight may
> result in an increase of RTT as well as in a decrease of RTT. This is due,
> as noted in many previous emails, to the fact that RTT variations may be
> due to many, many other reasons other than increasing or decreasing (by
> the sender) the number of packets in flight.
>
> > Phil
> >
> >
>
> ___________________________________________________________________
> Saad Biaz (Abou Youssef & Rim) Lecturer
> Texas A&M University Office : (409) 845-5007
> Department of Computer Science Home : (409) 862-9135
> College Station, TX, 77843-3112 Fax : (409) 847-8578
> ___________________________________________________________________
This archive was generated by hypermail 2b29 : Mon Jan 28 2002 - 09:12:20 EST