[Tmrg] TCP evaluation suite
rchertov at purdue.edu (Roman Chertov) Sat, 28 June 2008 19:10 UTC
From: "rchertov at purdue.edu"
Date: Sat, 28 Jun 2008 12:10:43 -0700
Subject: [Tmrg] TCP evaluation suite
In-Reply-To: <aa7d2c6d0806281142u70bfd47eu75099a323893195b@mail.gmail.com>
References: <aa7d2c6d0806271705g4f0363e7wd20787231f19ddac@mail.gmail.com> <48667F01.30005@purdue.edu> <aa7d2c6d0806281142u70bfd47eu75099a323893195b@mail.gmail.com>
Message-ID: <48668CB3.1050006@purdue.edu>
Lachlan Andrew wrote: > 2008/6/28 Roman Chertov <rchertov at purdue.edu>: >> Hello Lachlan, >> I think it would be worth while to include experiments which deal with >> admission of new flows into the current steady state. Such an experiment >> will allow to examine the impact of the startup stage on the already >> established flows. >> >> Roman > > Greetings Roman, > > Thanks for your input. I agree that we need to study arriving flows. > That was the goal of sections 4.3 and 4.4. Are you suggesting > changing them or adding something new? We currently don't consider > the impact of "slow start" on long-lived flows; is that what you are > suggesting? Yes, I think it would be valuable to look at scenarios where there is a collection of long-lived flows and a collection of short-lived flows. The short-lived flows send enough data for the slow start to increase the window several times, but not enough data to transition into congestion avoidance. This would be analogous to interleaving large and small file transfers. The obvious metrics to vary would be the arrival rate of short flows and the ratio of long-lived to short-lived flows. > > (I hope you don't mind my Cc'ing your good suggestion to the list.) Not a problem. Roman > > Cheers, > Lachlan >
- [Tmrg] TCP evaluation suite Lachlan Andrew
- [Tmrg] TCP evaluation suite Lachlan Andrew
- [Tmrg] TCP evaluation suite Roman Chertov
- [Tmrg] TCP evaluation suite Sally Floyd
- [Tmrg] TCP evaluation suite Fred Baker