[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [ANCP] Multicast Accounting



It thus seems that allowing for ANCP mcast accounting detail capability
negotiation seems as the way forward, with two main capabilities: simple
(start/stops only), detailed (start/stops with byte count). Arguably
time info can also be added, but appears to be implicit in the
start/stop info already.

-Woj.

> -----Original Message-----
> From: Derek Harkness [mailto:dharkness at juniper.net] 
> Sent: 17 January 2008 13:53
> To: Wojciech Dec (wdec); Stefaan DE CNODDER
> Cc: ancp at ietf.org
> Subject: RE: [ANCP] Multicast Accounting
> 
> I lost some of the the thread of the text alternatives, but 
> can only absolutely agree on the black box principle - AN & 
> NAS is one multicast node.
> 
> - Does anyone forsee a use for only simple start/stop mcast 
> ANCP accounting messages vs start/stop accounting with detailed info? 
> 
> There are multiple uses for this info:
> (1) Who is using what channels. This can help you with 
> network planning / overbooking, but also to enable targetted 
> advertising, for which you only may only need to know which 
> mcast channels they watch - i.e.
> start/stop.
> 
> (2) Charging for delivery of the mcast traffic - for which 
> you may want detailed info, as you are charging on volume 
> basis and time basis ( i.e.
> peak time volume will be more expensive ). This would apply for a e.g.
> wholesale network provider charging a retailer for delivery.
> 
> - And does anyone forsee a deployment model where a mix of 
> the modes is used, eg AN1 supporting detailed-mode and AN2 
> supporting not-detailed-mode? 
> 
> Less so, I think, unless you want the detailed mode from all 
> but the AN2 doesn't support it yet - so this is back to the 
> capabilities discussion i think. From an application scenario 
> this seems to make less sense.
> Accounting models tend to be application or network global 
> rather than network element or regionally specific, in my experience.
> 
> Cheers,
> 
> 	Derek.
> 
> +=============================+
> Derek Harkness
> Juniper Networks
> Nymphenburger Strasse 13-15
> 80335 Munich
> Germany
> Tel: +49 89 5529 4916
> Mobile: +49 172 843 6621
> Email: DHarkness at juniper.net
> +=============================+
> 
> -----Original Message-----
> From: Wojciech Dec (wdec) [mailto:wdec at cisco.com] 
> Sent: 14 January 2008 16:39
> To: Stefaan DE CNODDER
> Cc: ancp at ietf.org
> Subject: RE: [ANCP] Multicast Accounting
> 
> Stefaan,
> 
> I'll let Francois & Roberta comment on whether their proposal was to
> obly replace a bullet in Sec3.4.3, or both of them. In either case
> however, the proposal is very clear in removing the restrcition that
> mcast traffic accounting is out of scope, hence based on that reading
> the *current* framework text is not agreed to.
> 
> As for the philosphy behind ANCP, it's useful to keep in mind 
> the black
> box principle: ANCP should provide functionality to make the 
> combination
> of NAS + L2 Access-Node look like one black box. Hence, 
> irrespective of
> whether the AN can or cannot do unicat accounting, for the purpose of
> ANCP and multicast it appears rather useful to address the scenario
> where an AN *is* able to provide detailed mcast replication stats per
> port, beyond the basic case where it isn't.
> 
> In terms of capabilities, such detailed stat reports could 
> indeed be the
> subject of an additional negotiated capability, but on this 
> I'd like the
> WG to provide some comments. 
> - Does anyone forsee a use for only simple start/stop mcast ANCP
> accounting messages vs start/stop accounting with detailed info? 
> - And does anyone forsee a deployment model where a mix of 
> the modes is
> used, eg AN1 supporting detailed-mode and AN2 supporting
> not-detailed-mode?
> 
> Regards,
> Woj.
> 
> 
> > -----Original Message-----
> > From: Stefaan DE CNODDER 
> > [mailto:stefaan.de_cnodder at alcatel-lucent.be] 
> > Sent: 14 January 2008 14:56
> > To: Wojciech Dec (wdec)
> > Cc: ancp at ietf.org
> > Subject: Re: [ANCP] Multicast Accounting
> > 
> > 
> > Woj,
> > 
> >  From a functional point of view it is Ok but the comment was 
> > to change the second bullet of section 3.4.3 into what is 
> > already in the first bullet of section 3.4.3 (although there 
> > it is generalised that the AN can send this info to any acct 
> > server and not just the NAS).
> > 
> > My question is whether we can remove the first bullet.  If 
> > the AN has to do the multicast accounting, then it can also 
> > do the unicast accounting and I thought ANCP was about moving 
> > all of this to the NAS.  And if the AN can do the unicast 
> > accounting, then it can do the appropriate QoS queueing as 
> > well....  Anyhow, I am fine with the current text in the 
> > framework and if we believe that the comment below was on the 
> > wrong bullet, then it is all Ok.
> > 
> > I have more problems with the capabilities negotiated with ANCP. 
> > Implementation will not support all or nothing but that is 
> > what is negotiated in the capabilities.  Have the 
> > capabilities to be more fine-grained or should 
> > support/non-support be reflected in error codes?
> > 
> > regards,
> > Stefaan
> > 
> > 
> > Wojciech Dec (wdec) wrote:
> > 
> > > Stefaan,
> > > 
> > > then you're saying that you agree with Francois' proposal 
> > for Section 
> > > 3.4.3?
> > > 
> > > -Woj. 
> > > 
> > > 
> > >>-----Original Message-----
> > >>From: Stefaan DE CNODDER
> > >>[mailto:stefaan.de_cnodder at alcatel-lucent.be]
> > >>Sent: 14 January 2008 14:00
> > >>To: Wojciech Dec (wdec)
> > >>Cc: ancp at ietf.org
> > >>Subject: Re: [ANCP] Multicast Accounting
> > >>
> > >>
> > >>Woj,
> > >>
> > >>The framework draft currently does not exclude anything.  
> > >>Both options are possible.  The case where the AN is counting the 
> > >>packets and measuring the time is also in the framework and this 
> > >>information can be sent directly from the AN to the 
> > accounting server 
> > >>(which might be the NAS itself of course) with another 
> > protocol than 
> > >>ANCP like Radius for instance.
> > >>Actually, section 3.4.3 has to be read as whole and not 
> > just the part 
> > >>I copy/pasted in the email.  The question is basically if 
> > this section 
> > >>is Ok for everyone.
> > >>
> > >> From a protocol point of view, this means that each entry in the 
> > >>white list has the following information: indication whether 
> > >>accounting is needed or not, and if needed, whehter the 
> > accounting is 
> > >>performed by the NAS or by the AN.  In the latter case, 
> also the IP 
> > >>address of the accounting servers are specified.
> > >>
> > >>regards,
> > >>Stefaan
> > >>
> > >>
> > >>Wojciech Dec (wdec) wrote:
> > >>
> > >>>Hi Stefaan,
> > >>>
> > >>>it doesn't seem that we have agreement on this text, and an 
> > >>>alternative proposal is presented in 
> > >>>http://www1.ietf.org/mail-archive/web/ancp/current/msg00408.html.
> > >>>
> > >>>---snip---
> > >>>in Section "3.4.3 Multicast Accounting"
> > >>>
> > >>>
> > >>>replace:
> > >>>"
> > >>>o The AN keeps track of when replication starts or stops,
> > >>
> > >>and reports
> > >>
> > >>>this information to the NAS for further processing. In this
> > >>
> > >>case, ANCP
> > >>
> > >>>can be used to send the information from the AN to the NAS. 
> > >>
> > >>This can
> > >>
> > >>>be done with the Information Report message.
> > >>>The NAS can then generate the appropriate time and/or volume 
> > >>>accounting information per Access Loop and per multicast
> > >>
> > >>flow, to be
> > >>
> > >>>sent to the accounting system. The ANCP requirements to
> > >>
> > >>support this
> > >>
> > >>>approach are specified in this document.
> > >>>"
> > >>>by
> > >>>"
> > >>>o The AN keeps track of when replication for a given 
> > multicast flow 
> > >>>starts or stops and then the AN can convey time and/or volume 
> > >>>information to the NAS for further processing. This
> > >>
> > >>approach relies on
> > >>
> > >>>ANCP and is discussed below.
> > >>>
> > >>>---snip---
> > >>>
> > >>>In general for accounting to be of use it needs to be
> > >>
> > >>fairly accurate,
> > >>
> > >>>hence it is highly desirable for the access node to present
> > >>
> > >>to the NAS
> > >>
> > >>>the time and/or volume information for further 
> processing. Granted 
> > >>>that if access nodes are not capable of providing such 
> > information, 
> > >>>they will only signal simple start/stop events. I however 
> > don't see 
> > >>>why the framework and protocol should be limited only to
> > >>
> > >>this latter
> > >>
> > >>>variant. Is there a reason why the proposal presented by
> > >>
> > >>Francois is inadequate?
> > >>
> > >>>Regards,
> > >>>Woj.
> > >>>
> > >>>
> > >>>
> > >>>>-----Original Message-----
> > >>>>From: Stefaan DE CNODDER
> > >>>>[mailto:stefaan.de_cnodder at alcatel-lucent.be]
> > >>>>Sent: 08 January 2008 14:13
> > >>>>To: ancp at ietf.org
> > >>>>Subject: [ANCP] Multicast Accounting
> > >>>>
> > >>>>
> > >>>>Hi,
> > >>>>
> > >>>>At the last IETF meeting, there was a discussion about 
> multicast 
> > >>>>accounting is supposed to work.  Currently, the framework 
> > document 
> > >>>>mentions 2 options in section 3.4.3:
> > >>>>
> > >>>>---
> > >>>>   o  The AN keeps track of when replication starts or 
> stops, and
> > >>>>      generates the time and/or volume based accounting
> > >>
> > >>information
> > >>
> > >>>>per
> > >>>>      Access Loop and per multicast flow, before 
> sending it to a 
> > >>>>central
> > >>>>      accounting system for logging.  Since the AN
> > >>
> > >>communicates with
> > >>
> > >>>>      this accounting system directly, the approach
> > >>
> > >>doesn't require
> > >>
> > >>>>the
> > >>>>      use of ANCP.  It is therefore beyond the scope of this 
> > >>>>document;
> > >>>>
> > >>>>   o  The AN keeps track of when replication starts or 
> stops, and
> > >>>>      reports this information to the NAS for further
> > >>
> > >>processing.  In
> > >>
> > >>>>      this case, ANCP can be used to send the information
> > >>
> > >>from the AN
> > >>
> > >>>>to
> > >>>>      the NAS.  This can be done with the Information
> > >>
> > >>Report message.
> > >>
> > >>>>      The NAS can then generate the appropriate time 
> and/or volume
> > >>>>      accounting information per Access Loop and per
> > >>
> > >>multicast flow,
> > >>
> > >>>>to
> > >>>>      be sent to the accounting system.  The ANCP 
> requirements to
> > >>>>      support this approach are specified in this document.
> > >>>>---
> > >>>>
> > >>>>Is everyone Ok with this?  Is more needed, or less, or more 
> > >>>>clarrification on what it means?
> > >>>>
> > >>>>Personally, it looks Ok to me but I have a comment for extra 
> > >>>>clarrification on the first bullet: Does it require any
> > >>
> > >>impact on the
> > >>
> > >>>>ANCP protocol?  For the second bullet, I would expect that
> > >>
> > >>in a white
> > >>
> > >>>>list entry there is a flag indicating whether accounting is
> > >>
> > >>needed or
> > >>
> > >>>>not.  Is something similar needed to support the first 
> > bullet in a 
> > >>>>white list entry (e.g., need for time and/or volume based
> > >>
> > >>accounting
> > >>
> > >>>>with the IP address of acct server)?  Whatever is fine for
> > >>
> > >>me, but it
> > >>
> > >>>>has to be decided.
> > >>>>
> > >>>>regards,
> > >>>>Stefaan
> > >>>>
> > >>>>
> > >>>>
> > >>>>_______________________________________________
> > >>>>ANCP mailing list
> > >>>>ANCP at ietf.org
> > >>>>https://www1.ietf.org/mailman/listinfo/ancp
> > >>>
> > >>>
> > >>
> > >>_______________________________________________
> > >>ANCP mailing list
> > >>ANCP at ietf.org
> > >>https://www1.ietf.org/mailman/listinfo/ancp
> > > 
> > > 
> > 
> 
> 
> _______________________________________________
> ANCP mailing list
> ANCP at ietf.org
> https://www1.ietf.org/mailman/listinfo/ancp
> 


_______________________________________________
ANCP mailing list
ANCP at ietf.org
https://www1.ietf.org/mailman/listinfo/ancp