[183393] in North American Network Operators' Group
Re: Zero rating implentation strategies
daemon@ATHENA.MIT.EDU (Christian Kuhtz)
Tue Sep 1 13:37:01 2015
X-Original-To: Nanog@nanog.org
From: Christian Kuhtz <chkuhtz@microsoft.com>
To: Jean-Francois Mezei <jfmezei_nanog@vaxination.ca>
Date: Tue, 1 Sep 2015 17:35:52 +0000
In-Reply-To: <55E5DC9A.5050108@vaxination.ca>
Cc: "Nanog@nanog.org" <Nanog@nanog.org>
Errors-To: nanog-bounces@nanog.org
Inline..
On Sep 1, 2015, at 10:13 AM, Jean-Francois Mezei <jfmezei_nanog@vaxination.=
ca<mailto:jfmezei_nanog@vaxination.ca>> wrote:
On 15-09-01 12:36, Christian Kuhtz wrote:
Zero rating is not a new concept. It has existed in the mobile world since =
the days of the dumb phone.
Yes, but is now in a different scale and when you start to zero rate
content that comes from CDN (aka: many IPs which may also serve non zero
rated content).
I don't agree with that assertion. Not a new concept. I worked on stuff li=
ke that over a decade ago, with multiple CDN providers. Wasn't even mobil=
e, it was for DSL.
There are scaling issues if the idea was out of touch with how interwebs ac=
tually work. Only in those cases might it have technically worked but pric=
ed itself out of reality. But there wasn't anything fundamentally to say i=
t couldn't be done.
CDN's have for a very long time had ways to implement custom mappings and t=
weak their algorithms to source specific content to a specific subscriber i=
n a very specific way or place source nodes in particular locations (logica=
lly or physically). They only varied in their ability or willingness to do=
so on terms that might seem reasonable to a given partner (which generally=
means you either do or don't come to terms on contracts). It wasn't a fun=
damental technical barrier even 10-15 yrs ago.
Got a reference to why this was killed by the regulator in Canada?
Bell Mobility zero rating its own MobileTV offering was decided in
January 2015 by CRTC:
Thank you. The key here appears to be this:
Subsection 27(2) prohibits Canadian carriers from
conferring an undue disadvantage to others, or an undue preference to
itself or others.
So this appears to be about distorting the competition between their own se=
rvices and third parties rather than about zero rating per se.
Mobile networks typically use their packet core (and prior iterations of th=
e same termination, rating, billing, management gear) to rate and bill spec=
ific to each subscriber. It is done with voice minutes, text, data. Wheth=
er or not you consider the solutions scalable is up to one's individual jud=
gement. But this zero rating billing model has and is very widely deployed =
and at massive scale. In fact, there are specific interfaces defined for ju=
st these purposes in the applicable standards. There are many many conceiv=
able ways in which billing mediation and associated infrastructure for zero=
rating can and is implemented. This is not new and is very well understood=
in the mobile industry.
Not sure what you're after here.
What I am after is what sort of logic is used to determine whether a
packet is to be counted towards total monthly usage or not. Is it based
on sourse IP address ? DPI equipment looking at content ? And more
importantly, if it is possible to zero rate any flow that is below a
certain speed. (whether from a radio station or some heart monitor).
All of the above and more. Depends on content. For example, if you obtain =
the knowledge of origin and subscriber, you have what is needed. Origin sho=
uld not be constrained to origin in terms of IP address space or port numbe=
rs. It can be marked in the protocol header, content, sourced from specific=
interfaces etc etc. The ways are pretty endless and the gear is readily av=
ailable. There is an entire cottage industry peddling this stuff from sourc=
ing, detection through rating and billing in addition to the major equipmen=
t vendors.
There are no inherent limits. You can do it with or without DPI. You can do=
it with or without subscriber management systems (like BRAS). You can do i=
t with or without knowledge of source or destination IP addresses. You can =
do it with any kind of content. You can do it with or without flow awarenes=
s. You can do it with or without special protocol semantics.
What is capable is (as always) up to the engineering clue, business savvy a=
nd business model supporting the implementation. And I really mean anythin=
g can, has, and will be brought to life at scale given a suitable business =
cause.
I don't mean to be obnoxious. The question is a bit like asking if interweb=
s have limits ;-).. Yes, there are stupid ways to screw things up, but ther=
e is a huge variety of ways to make it work and make money doing it. And wi=
thout pissing off regulators. There is no red flag from a regulatory persp=
ective based on the technology itself in my opinion.
Hope this helps.
Best regards,
Christian
