[195327] in North American Network Operators' Group
RE: ISP billing - data collection, correlation, and billing
daemon@ATHENA.MIT.EDU (Sean Pedersen)
Mon Jul 17 12:17:28 2017
X-Original-To: nanog@nanog.org
From: "Sean Pedersen" <spedersen.lists@gmail.com>
To: "'Michael Krygeris'" <me@krygerism.com>, "'Lee Howard'" <lee@asgard.org>,
"'Luke Guillory'" <lguillory@reservetele.com>, <nanog@nanog.org>
In-Reply-To: <CAAqp_xvAPSj4r1Le=BzkWxacsigD2T0Aq69ANpivhwy9LfSnag@mail.gmail.com>
Date: Mon, 17 Jul 2017 09:17:23 -0700
Errors-To: nanog-bounces@nanog.org
Flexible NetFlow is the best option for us. I haven=E2=80=99t seen a lot =
of products outside of open source that work with telemetry at the =
moment, and nothing that has billing functions. We=E2=80=99re far too =
lean to be able to rely on open source / devops at this point. =
That=E2=80=99s been a hard lesson to teach the company.=20
=20
It=E2=80=99s definitely on our internal roadmap, though.
=20
From: Michael Krygeris [mailto:me@krygerism.com]=20
Sent: Sunday, July 16, 2017 9:35 AM
To: Lee Howard <lee@asgard.org>; Luke Guillory =
<lguillory@reservetele.com>; Sean Pedersen <spedersen.lists@gmail.com>; =
nanog@nanog.org
Subject: Re: ISP billing - data collection, correlation, and billing
=20
=20
On Sat, Jul 15, 2017 at 4:48 AM Lee Howard <lee@asgard.org =
<mailto:lee@asgard.org> > wrote:
=20
Depending on the capability of the exporter, you don't need to export =
full flow information. With Cisco's Flexible Netflow you can define the =
aggregation in the flow cache you are monitoring. You are not required =
to use a 7-tuple.
An aggregation could be something basic like this:
Source interface
Destination interface=20
Octets
Packets
=20
This would give you SNMP equivalent for byte accounting on interfaces =
without requiring full flow accounting IF you're not forced to do =
sampling and IF you have flexible netflow.
=20
Another much more recent method around SNMP (sorry SNMP, I'm over you) =
is streaming telemetry, which is part of Netconf/YANG/OpenConfig.
This is more of a push method for these yang data models(the relevent =
one here being snmp interfaces table).
It already exists on some Juniper and Cisco platforms.
Mike Krygeris
=20
On 7/14/17, 2:47 PM, "NANOG on behalf of Luke Guillory"
<nanog-bounces@nanog.org <mailto:nanog-bounces@nanog.org> on behalf of =
lguillory@reservetele.com <mailto:lguillory@reservetele.com> > wrote:
>On the HFC / CMTS side of things we have IPDR which I believe has some
>open source collectors out there. I'm not sure that IPDR is used much
>outside of the HFC world though.
IPDR was my first thought as an alternative to SNMP. Is its accuracy
comparable? It=E2=80=99s been included into TR-069, so it=E2=80=99s =
theoretically
available to telcos, too. And usage-based billing is part of =
it=E2=80=99s purpose.
Not sure I=E2=80=99d want to use NetFlow/IPFIX, since by nature it =
tracks flows,
not bits, and I don=E2=80=99t want to record flows. But I=E2=80=99d be =
interested in
hearing others=E2=80=99 experience.
Lee
