On 2/11/14, 1:43 AM, "Mikael Abrahamsson" <swm...@swm.pp.se> wrote:
>On Mon, 10 Feb 2014, Greg White wrote: > >> Also, we model shared link congestion by modulating the capacity >> available to our tested user. This was described in our April 2013 >> paper. Since one user's queuing latency doesn't impact another user, >>the >> interaction between users is largely limited to how much capacity is >> available to our tested user at any point in time. Our model was, > >Could you please elaborate on this. If we're 10 users all using our 50 >megs up, now 5 of them stop sending, since now the upstream system >capacity per actually sending user has doubled, how can the queueing >delay >not be affected? What happens when those 5 that now went silent started >sending again? It is just as you think it is. When 10 users are placing equal, saturating load on the link, they will each get (on average) 10% of the link rate. When there are 5 users, they will each get (on average) 20% of the link rate. This is enforced by the CMTS. So when the active users undergoes a step change from 10 to 5, each of the 5 remaining active users will see a doubling of their link rate. The reverse will happen when the active users returns to 10. Step changes in link rate will result in step changes in queuing latency. This we model. What I was getting at is that the above is true for user #1 regardless of what applications and what AQM user #2 is using. So we don't need to simulate user #2 (it is sufficient to model him) if we're interested in user #1's performance. -Greg > >What I'm after is to measure the system as a whole, the CPE queueing >algorithm together with the cable "system" interaction and how this would >interact with TCP streams. > >The same use-case would be valid for PON. > >-- >Mikael Abrahamsson email: swm...@swm.pp.se _______________________________________________ aqm mailing list aqm@ietf.org https://www.ietf.org/mailman/listinfo/aqm