[171881] in North American Network Operators' Group
Re: Observations of an Internet Middleman (Level3)
daemon@ATHENA.MIT.EDU (Livingood, Jason)
Thu May 15 17:20:18 2014
X-Original-To: nanog@nanog.org
From: "Livingood, Jason" <Jason_Livingood@cable.comcast.com>
To: Scott Berkman <scott@sberkman.net>, "nanog@nanog.org" <nanog@nanog.org>
Date: Thu, 15 May 2014 21:19:22 +0000
In-Reply-To: <5375209A.20003@sberkman.net>
Errors-To: nanog-bounces@nanog.org
On 5/15/14, 4:16 PM, "Scott Berkman" <scott@sberkman.net<mailto:scott@sberk=
man.net>> wrote:
Everyone knows Comcast uses (or used) Sandvine for shaping (unless
they've finished building a new probably internal solution, I'm sure
this is another secret we'll only have rumors to work with, ).
Comcast turned off Sandvine=92s active traffic management system at the end=
of 2008; I know because it was my job to do it (and I had nothing to do wi=
th the decision to turn on the Sandvine system). ;-) FCC Chairman Kevin Mar=
tin required the turn down of that system by EOY 2008. Here is the letter t=
o the FCC confirming that transition was completed on January 9, 2009: http=
://downloads.comcast.net/docs/comcast-nm-transition-notification.pdf.
It was replaced with a protocol-agnosting congestion management system (act=
ive only in the DOCSIS network). That system was disclosed here:
http://downloads.comcast.net/docs/Attachment_B_Future_Practices.pdf
http://tools.ietf.org/html/rfc6057
http://downloads.comcast.net/docs/IETF%2072%20-%20TANA%20BoF%20-%20ISP%20Re=
quirements%20-%20Comcast.pdf
http://downloads.comcast.net/docs/Comcast-IETF-P2Pi-20080528.pdf
http://downloads.comcast.net/docs/ietf-p2pi-comcast-20080509.pdf
There is no other active traffic management system (other than what any ISP=
has for DDoS protection/mitigation), period.
I'm no longer in a position to test this for reporting to the FCC as
suggested, but in a previous life we were able to prove it enough for the C=
omcast customer getting the short end of the stick to stop yelling at us an=
d get a new provider, which of course made everyone involved happier.
We used to have a =93positive=94 traffic shaping system called PowerBoost. =
That enabled customers to boost above their advertised or provisioned rates=
for brief periods. That system seemed to cause more customer confusion tha=
n it was worth and PowerBoost was eliminated across all of our tiers of ser=
vice. Sometimes tools to notice traffic shaping noticed PowerBoost and it w=
as sometimes hard to explain that we were shaping traffic *up* in capacity =
rather than down, but I digress.
I'd love to see a case study published by
Comcast on how that project went and what the impacts to the network and bo=
ttom line were.
We documented every step of the way on our Network Management page at http:=
//networkmanagement.comcast.net/. You may also be interested to read Alissa=
Cooper=92s September 2013 PhD thesis, which touches on this system on some=
level at http://www.alissacooper.com/files/Thesis.pdf.
There is also a good paper by the Broadband Internet Technical Advisory Gro=
up (BITAG) on this topic at http://www.bitag.org/documents/BITAG_-_Congesti=
on_Management_Report.pdf. We at Comcast comply with all of the BITAG recomm=
endations in that paper.
Jason
