From mboxrd@z Thu Jan 1 00:00:00 1970 From: Gerrit Renker Date: Mon, 05 Mar 2007 09:49:49 +0000 Subject: Re: [dccp] Question on resetting nominal send time Message-Id: <200703050949.49387@strip-the-willow> List-Id: References: <5640c7e00701161351q19360335mf2862a31596ae94a@mail.gmail.com> In-Reply-To: <5640c7e00701161351q19360335mf2862a31596ae94a@mail.gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable To: dccp@vger.kernel.org Hi Eddie, Sorry for the delay in responding. | What follows is a first cut at a solution. =A0Any thoughts from others?? | =20 | If t_ipi is used to schedule transmissions, then the following equation = should=20 | be applied each time the application is scheduled: | =20 | =A0 =A0 t_ipi :=3D max(t_ipi, t_now - RTT/2) | =20 | This never lets t_ipi fall more than 1/2 RTT behind the current time. = =A0An=20 | application is still allowed to send packets in a small burst after an i= dle=20 | period, but the size of that burst is limited to RTT/2 worth of packets. | =20 | RTT/2 was chosen because senders can send 2*last_receive_rate in any RTT. | =20 | I am sure that this simple choice has disadvantages, such as little burs= ts at=20 | the beginnings of idle periods. =A0One could be more conservative and se= t e.g. | =20 | =A0 =A0 =A0t_ipi :=3D max(t_ipi, t_now - t_gran). | =20 | But I think RTT/2 might be OK. =A0Implementation experience would be pre= ferred. | =20 | This issue is really an implementation issue. =A0RFC3448 4.6 is not exac= tly=20 | normative; it discusses one way to achieve a send rate, not a required=20 | implementation. =A0So in some sense the implementer is free to choose an= ything=20 | reasonable. In TFRC t_ipi is always smaller than RTT, so RTT/2 is an upper bound. I thi= nk it makes sense (from an implementation standpoint) to use one full t_ipi as upper bo= und. This is similar to your solution in that both values are less than RTT, and both= provide a means to stop large `packet storms'.=20 The reason for chosing t_ipi is that the size of the large burst depends on= the number of full t_ipi intervals that fit into the time interval that the receiver i= s lagging=20 behind (can send detailed derivation). But, as said, both choices are simil= ar. Gerrit