[171379] in North American Network Operators' Group
RE: The FCC is planning new net neutrality rules. And they could
daemon@ATHENA.MIT.EDU (Hugo Slabbert)
Sun Apr 27 23:54:05 2014
X-Original-To: nanog@nanog.org
From: Hugo Slabbert <hslabbert@stargate.ca>
To: "nanog@nanog.org" <nanog@nanog.org>
Date: Mon, 28 Apr 2014 03:52:19 +0000
In-Reply-To: <CAObxEFDCDGoMMnjoALsg7Jcev5UccstTVMvfiwWt=KtGQkR4-g@mail.gmail.com>
Errors-To: nanog-bounces@nanog.org
Apologies that I dropped offlist as I was out for the day. I think the bul=
k of my thoughts on this have already been covered by others since, includi=
ng e.g. Matt's poor grandmother and her phone dilemma in the "What Net Neut=
rality should and should not cover" thread.
Basically I think we're on the same page for the most part, with maybe some=
misunderstandings between us.
> I covered this scenario in more detail in my post "What Net Neutrality sh=
ould and should not cover" but if you expand on the assumption that paying =
for an internet connection also pays for the direct connection of every par=
ty who you exchange traffic with then you have a scenario where only half t=
he people connected to the Internet should have to pay at all for their con=
nection because any scenario where people simply buy their own pipe would b=
e considered "double billing".
I don't think anyone on the Netflix^H^H^H^H^H^H $ContentProvider side of th=
is was saying that $ContentProvider should get everything handed to them on=
a silver platter. $ContentProvider pays for transit sufficient to handle =
the traffic that their customers request. $EyeballNetwork's customers pay =
it for internet access, i.e. to deliver the content that they request, e.g.=
from $ContentProvider. That covers both directions here. Links between $=
ContentProvider's transit provider and $EyeballNetwork were getting congest=
ed, and $EyeballNetwork refuses to upgrade capacity. Where we were getting=
into the double-dip was $EyeballNetwork saying to $ContentProvide: "Hey, =
we know you already pay for transit, but you're gonna have to pay us as wel=
l if you want us to actually accept the traffic our customers requested".
The alternate arrangement between $ContentProvider and $EyeballNetwork seem=
s to be private peering, where again it would seem to be fair for each side=
to bring the needed transport and ports to peering points. In recent hist=
ory, though, it seems that $EyeballNetwork came out ahead in that agreement=
somehow. Now, Tore brought up a good point on paid peering in cases where=
e.g. $EyeballNetwork is already exchanging traffic with $ContentProvider t=
hrough existing peering or below their CDR on existing transit, and indeed =
it seems that was the case for $EyeballNetwork via peering with $CheapTrans=
itProvider that $ContentProvider was using. But it seems that $EyeballNetw=
ork was having a pissing match with $CheapTransitProvider and refusing to u=
pgrade ports.
"Okay", says $ContentProvider. "How about we just peer directly."
"Sounds great," says $EyeballNetwork. "Since we have to allocate capacity =
for this discrete from our existing peering capacity, you'll need to foot t=
he bill for that."
"Huh?" says $ContentProvider. "This could have been fixed by you increasin=
g your peering capacity to match the traffic volume your users are requesti=
ng, but you didn't want to do that because of your tiff with $CheapTransitP=
rovider. Tell me again why we're paying for your side of this *in addition=
* to our own when we're only going this route because of a decision *you* m=
ade?"
"Because you need to reach our customers, and we're the only path to them, =
so we have leverage."
*blank stare*
"So you're willing to give your customers crappy service because your custo=
mers don't have alternate options and you think we need this more than you =
do?"
"That's a possibility."
"I hate you."
"I know; sign here please."
But, again, this is outside looking in. For now, I'll pick up a copy of Bi=
ll Norton's Internet Peering book as per Bob's suggestion, for some light S=
unday night reading.
Cheers,
--
Hugo
________________________________
From: Rick Astley <jnanog@gmail.com>
Sent: Sunday, April 27, 2014 8:45 AM
To: Hugo Slabbert
Cc: nanog@nanog.org
Subject: Re: The FCC is planning new net neutrality rules. And they could e=
nshrine pay-for-play. - The Washington Post
If it were through a switch at the exchange it would be on each of them to =
individually upgrade their capacity to it but at the capacities they are at=
it they are beyond what would make sense financially to go over an exchang=
e switch so they would connect directly instead. It's likely more along the=
lines of needing several 100G ports as Netflix is over 30% of peak usage t=
raffic in North America:
"Netflix (31.6%) holds its ground as the leading downstream application in =
North America and together with YouTube (18.6%) accounts for over 50% of do=
wnstream traffic on fixed networks." (source https://www.sandvine.com/tren=
ds/global-internet-phenomena/ )
That amount of data is massive scale. I don't see it as double dipping beca=
use each party is buying the pipe they are using. I am buying a 15Mbps pipe=
to my home but just because we are communicating over the Internet doesn't=
mean the money I am paying covers the cost of your connection too. You mus=
t still buy your own pipe in the same way Netflix would. I covered this sce=
nario in more detail in my post "What Net Neutrality should and should not =
cover" but if you expand on the assumption that paying for an internet conn=
ection also pays for the direct connection of every party who you exchange =
traffic with then you have a scenario where only half the people connected =
to the Internet should have to pay at all for their connection because any =
scenario where people simply buy their own pipe would be considered "double=
billing".
The cost for residential broadband is high enough in the US without a polic=
y like that in place. If there is one policy that would keep poor families =
from being able to afford broadband it would be that one.
On Sun, Apr 27, 2014 at 2:58 AM, Hugo Slabbert <hslabbert@stargate.ca<mailt=
o:hslabbert@stargate.ca>> wrote:
> ...but if that point of congestion is the links between Netflix and Comca=
st...
Which, from the outside, does appear to have been the case.
> ...then Netflix would be on the hook to ensure they have enough capacity =
to Comcast to get the data at least gets TO the Comcast network.
Which I don't believe was a problem? Again, outside looking in, but the ap=
pearances seemed to indicate that Comcast was refusing to upgrade capacity/=
ports, whereas I didn't see anything indicating that Netflix was doing the =
same. So:
> I have gear; you have gear. I upgrade or add ports on my side; you upgra=
de or add ports on your side.
> The argument at hand is if Comcast permitted to charge them for the links=
to get to their network or should they be free/settlement free. I think it=
should be OK to charge for those links as long as its a fair market rate a=
nd the price doesn't basically amount to extortion.
Are we talking here about transport between Netflix's POPs and Comcast's? =
I definitely don't expect Comcast to foot the bill for transport between th=
e two, and if Netflix was asking for that I'm with you that would be out of=
line. If there are existing exchange points, though, would it not be reas=
onable to expect each side to up their capacity at those points?
> Once that traffic is given directly to comcast no other party receives pa=
yment for delivering it so there is no double billing.
The "double-dip" reference was to charging both the content provider and th=
e ISP's own customer to deliver the same bits. If the traffic from Netflix=
was via Netflix's transit provider and Comcast then again was looking to b=
ill Netflix to accept the traffic, we'd hit double billing.
I guess that's the question here: If additional transport directly been PO=
Ps of the two parties was needed, somebody has to pay for the links. Relea=
ses around the deal seemed to indicate that the peering was happening at IX=
s (haven't checked this thoroughly), so at that point it would seem reasona=
ble for each party to handle their own capacity to the peering points and c=
all it even. No?
--
Hugo
________________________________
From: Rick Astley <jnanog@gmail.com<mailto:jnanog@gmail.com>>
Sent: Saturday, April 26, 2014 11:23 PM
To: Hugo Slabbert
Cc: nanog@nanog.org<mailto:nanog@nanog.org>
Subject: Re: The FCC is planning new net neutrality rules. And they could e=
nshrine pay-for-play. - The Washington Post
>How is this *not* Comcast's problem? If my users are requesting more traf=
fic than I banked on, how is it not my responsibility to ensure I have capa=
city to handle that? I have gear; you have gear. I upgrade or add ports o=
n my side; you upgrade or add ports on your side. Am I missing something?
Sort of yes, it's Comcasts problem to upgrade subscriber lines but if that =
point of congestion is the links between Netflix and Comcast then Netflix w=
ould be on the hook to ensure they have enough capacity to Comcast to get t=
he data at least gets TO the Comcast network. The argument at hand is if Co=
mcast permitted to charge them for the links to get to their network or sho=
uld they be free/settlement free. I think it should be OK to charge for tho=
se links as long as its a fair market rate and the price doesn't basically =
amount to extortion. Sadly the numbers are not public so I couldn't tell yo=
u one way or the other aside from I disagree with the position Netflix seem=
s to be taking that they simply must be free. Once that traffic is given di=
rectly to comcast no other party receives payment for delivering it so ther=
e is no double billing.
This diagram best describes the relationship (ignoring pricing): http://www=
.digitalsociety.org/files/gou/free-and-paid-peering.png
"Content provider" would be Netflix and Comcast would be Broadband ISP 1.
On Sun, Apr 27, 2014 at 1:56 AM, Hugo Slabbert <hslabbert@stargate.ca<mailt=
o:hslabbert@stargate.ca><mailto:hslabbert@stargate.ca<mailto:hslabbert@star=
gate.ca>>> wrote:
Okay, I'm not as seasoned as a big chunk of this list, but please correct m=
e if I'm wrong in finding this article a crock of crap. With Comcast/Netfl=
ix being in the mix and by association Cogent in the background of that the=
re's obviously room for some heated opinions, but here goes anyway...
>A long, long time ago when the Internet was young and few, if any had thou=
ght
>to make a profit off it, an unofficial system developed among the network
>providers who carried the traffic: You carry my traffic and I'll carry you=
rs
>and we don't need money to change hands. This system has collapsed under
>modern realities.
I wasn't aware that settlement-free peering had "collapsed". Not saying it=
's the "only way", but "she ain't dead yet".
Seltzer uses that to set up balanced ratios as the secret sauce that makes =
settlement-free peering viable:
"The old system made sense when the amount of traffic each network was send=
ing to the other was roughly equivalent."
...and since Netflix sends Comcast more than it gets, therefor Netflix need=
s to buck up:
"Of course Netflix should pay network providers in order to get the huge am=
ounts of bandwidth they require in order to reach their customers with suff=
icient quality."
But this isn't talking about transit; this is about Comcast as an edge netw=
ork in this context and Netflix as a content provider sending to Comcast us=
ers the traffic that they requested. Is there really anything more nuanced=
here than:
1. Comcast sells connectivity to their end users and sizes their network a=
ccording to an oversubscription ratio they're happy with. (Nothing wrong h=
ere; oversubscription is a fact of life).
2. Bandwidth-heavy applications like Netflix enter the market.
3. Comcast's customers start using these bandwidth-heavy applications and =
suck in more data than Comcast was betting on.
4. Comcast has to upgrade connectivity, e.g. at peering points with the he=
avy inbound traffic sources, accordingly in order to satisfy their customer=
s' usage.
How is this *not* Comcast's problem? If my users are requesting more traff=
ic than I banked on, how is it not my responsibility to ensure I have capac=
ity to handle that? I have gear; you have gear. I upgrade or add ports on=
my side; you upgrade or add ports on your side. Am I missing something?
Overall it seems like a bad (and very public) precedent & shift towards dou=
ble dipping, and the pay-for-play bits in the bastardized "Open Internet" r=
ules don't help on that front. Now, Comcast is free to leverage their cust=
omers as bargaining chips to try to extract payments, and Randy's line of e=
ncouraging his competitors to do this sort thing seems fitting here. Basic=
ally this doesn't harm me directly at this point. Considering the lack of =
broadband options for large parts of the US, though, it seems that end user=
s are getting the short end of the stick without any real recourse while th=
at plays out.
--
Hugo
________________________________________
From: NANOG <nanog-bounces@nanog.org<mailto:nanog-bounces@nanog.org><mailto=
:nanog-bounces@nanog.org<mailto:nanog-bounces@nanog.org>>> on behalf of Lar=
ry Sheldon <LarrySheldon@cox.net<mailto:LarrySheldon@cox.net><mailto:LarryS=
heldon@cox.net<mailto:LarrySheldon@cox.net>>>
Sent: Saturday, April 26, 2014 4:58 PM
To: nanog@nanog.org<mailto:nanog@nanog.org><mailto:nanog@nanog.org<mailto:n=
anog@nanog.org>>
Subject: Re: The FCC is planning new net neutrality rules. And they could e=
nshrine pay-for-play. - The Washington Post
h/t Suresh Ramasubramanian
FCC throws in the towel on net neutrality
http://www.zdnet.com/fcc-throws-in-the-towel-on-net-neutrality-7000028770/
Forward! On to the next windmill, Sancho!
--
Requiescas in pace o email Two identifying characteristics
of System Administrators:
Ex turpi causa non oritur actio Infallibility, and the ability to
learn from their mistakes.
(Adapted from Stephen Pinker)
