Comcast's Protocol-Agnostic Congestion Management System
RFC 6057

Note: This ballot was opened for revision 09 and is now closed.

(Lars Eggert) Yes

(Jari Arkko) No Objection

Comment (2010-08-25 for -** No value found for 'p.get_dochistory.rev' **)
No email
send info
A review by Ari Keranen:


There's some repetition especially in section 7. For example, the following sentence is there twice:

   As a
   result, the congestion management system measures the traffic
   conditions as observed by each CMTS, and applies any policy actions
   to traffic on a specific interface of a CMTS (rather than some other,
   more distant segment of the network).


The described approach may have at least one problem: given that there is a "fixed" limit (e.g., 70% utilization for 15 minutes) after which user enters the Extended High Consumption State (EHCS), it gives users incentives to linger just below the limit (e.g., 100% utilization for 14.5 minutes and 60% for 30 seconds). Even if the limits would be unknown to subscribers, the user could try to probe them and try to stay just below the probed limit and thus get a bigger share of the congested link than users not doing that.

Instead, if the user could be partially in ECHS, depending on the utilization after some threshold, there would not be such incentives. For example, after using 80% of the link for 15 minutes (or 100% for 5 minutes), one third of the packets would be marked BE, or for 90% utilization for 15 minutes, two thirds of the packets would be BE, etc. This approach would also prevent the oscillation problem.

In section 7.3 it is mentioned that (in theory) BE traffic might not be delivered at all in case of high link utilization. Wouldn't it be better to deliver BE packets just with (much) lower probability?

(Ron Bonica) No Objection

(Stewart Bryant) No Objection

(Adrian Farrel) No Objection

(David Harrington) No Objection

Comment (2010-08-23 for -** No value found for 'p.get_dochistory.rev' **)
No email
send info
in 3.7, any device controlled by Comast? or any device?

(Russ Housley) No Objection

(Robert Sparks) No Objection

(Sean Turner) No Objection