[e2e] RTT estimation

[Alhussein Abouzeid]

You estimate RTT in order to set the RTO. I think you are looking for this:

S.W. Edge, "An adaptive timeout algorithm for retransmission across a packet
switching network," Computer Communications Review vol. 14, no. 2, p.248-55,
1984.

The paper is not online (as far as I could tell). While much work has been
done seen then,  it is worth a quick visit to the library since this paper
includes an analysis of the adaptive (exp. weighted) timeout mechanism
(specifically, as I recall, the delay subject to a given false timeout
probability). The analysis is not limited to a specific delay distribution,
but, as far as I recall, it assumes a stationary delay process and requires
certain conditions for the case of non-stationary delay.

Hope this helps.

Hussein
--www.ecse.rpi.edu/~abouza



----- Original Message -----
From: "Dong Zheng" <Dong.Zheng at asu.edu>
To: <end2end-interest at postel.org>
Sent: Wednesday, April 30, 2003 2:29 PM
Subject: [e2e] RTT estimation


> Hi,
>
> Sorry to bother you all, I have a simple question to ask.
>
> We know that TCP keeps a dynamic RTT estimation and uses this
> estimation to set the timer for congestion avoidance. The estimation
> keep track of the RTT fluctuation and hence it will affect the
> congestion avoidance behavior. If the estimation is not good, it will
> trigger unnecessary retransmissions.
>
> My question is whether there is any metric that tells how well the
> current Jacobson/Karn algorithm doing the estimation? I mean, given the
> fluctuation of the RTT, to what level of the fluctuation can the current
> algorithm track?
>
> Thank you very much.
>
> -Dong
>