Re: [Lsr] Congestion (flow) control thoughts.

["Gengxuesong (Geng Xuesong)" <[e-mail]>]

Hi,



Thank you Chris for emphasizing that congestion control is hard and more reference may be necessary.

And as the the follow-up of my comments in the interim meeting, I would like to make my question more clear:

In congestion control of layer 4, it is assumed that there is a bottleneck in the network, and the ideal rate of the transmitters equals to a fair share of the bandwidth in the bottleneck. The flows in the network change all the time and so as to the ideal transmitting rate. There are some methods to detect the bottleneck, for example detecting packet loss, setting ECN, RTT and so on. Considering that the goal of cc is to maximize the throughput and minimize the queuing delay, the throughput and delay could be used to compare different cc algorithms.

The problem of mine is that for CC of ISIS flooding, where is the bottleneck?(maybe the receiver capability) What method of detecting could be used? (In my understanding, we have two options at this stage: one from the receiver side and the other from the transmitter side) What is the criteria of comparing? (still a little confusing for me )



Best Regards

Xuesong



-----Original Message-----

From: Lsr [mailto:[e-mail]] On Behalf Of Christian Hopps

Sent: Wednesday, April 29, 2020 8:24 PM

To: [e-mail]

Cc: [e-mail]

Subject: [Lsr] Congestion (flow) control thoughts.



[As WG member]



I like the start these drafts are making, I have a few thoughts.



- I think its worthwhile work and as a WG we should end up adopting a draft or drafts that try to solve the problem.

- Congestion control is hard, and a very well studied problem, as such whatever we end up should reference prior-art to justify it's claims.

- Data, whatever solution we end up with needs to have data to back up it's claims.



** draft-ginsberg-lsr-isis-flooding-scale



Having just recently offered a, "CC algorithm is local so we aren't specifying it", in some unrelated work, I discovered (correctly so) that it was not sufficient (see [*] for the resulting changes if curious). One needs to provide (doesn't need to be normative) a working algorithm to prove that the protocol is providing enough information to the transmitter to actually implement a viable CC algorithm.



Now draft-ginsberg does suggest an algorithm, however only for P2P. We need to address LAN as well. This also needs references on how this algorithm relates to existing studied work, and perhaps most importantly some data showing that it actually works.



Without this it will almost for sure not clear the IESG (in particular Transport AD review), and rightly so. :)



** draft-decraene-lsr-isis-flooding-speed-03



I read this draft as providing feedback to transmitters on ways to modify the parameters of their CC algorithm. It doesn't actually seem in conflict with draft-ginsberg (or doesn't need to be) to me. I believe it's following the path of least resistance in utilizing existing IS-IS parameters. I wonder if we could be more adventurous here.



** Both



Both drafts mention trying to increase the ACK rate. I think we need to look deeply into this, as this is critical information for CC algorithms. It very well could be that we do not need protocol changes as the protocol provides an ACK (or NACK for LANs), but I also don't think changes or additions should be verboten either if they could provide significant improvements.



In the algorithms we propose or envision we need to be considering the affect of larger RTT on the algorithm for slow or long distant links (radios, satellites, long-haul fibers, etc).



Thanks,

Chris.

[As WG member]



[*] - https://tools.ietf.org/html/draft-hopps-ipsecme-iptfs-01#section-3

      https://tools.ietf.org/html/draft-hopps-ipsecme-iptfs-01#appendix-B),