RE: [Diffserv] CountActTable

["Kwok-Ho Chan" <khchan@nortelnetworks.com>]

I have posted the following on the DiffServ list also.
Please skip if you read it on the DiffServ list first.

And I am glad we have agreements.

Yes, currently work is under way for a QoS Accounting Policy PIB 
(not exact or finalized name, so please don't comment on the name :)).
separate from the Accounting Framework PIB. 
This will include capability of where statistics can be gathered
(different 
device put counters in different ASICs of different QoS datapaths), 
where statistics need to be gathered (as determined by the PDP), 
and what statistics need to be gathered. 
Please note the statistics discussed here are relavant to a specific
policy instance at the PEP. 
Please also note this PIB does not intended to indicate how the
statistics 
will be used. Billing and other applications are considered application
layers above the PDP that does not need to be indicated in the
information 
exchanged between the PDP and PEP. 
Hope this clear some of air up. 
-- Kwok Ho Chan --


At 08:35 PM 4/5/01 +0000, Iliff, Tina wrote:

Agree.
Tina Iliff

-----Original Message----- From:   Moisand, Jerome [mailto:jmoisand@unispherenetworks.com] Sent:   Thursday, April 05, 2001 3:23 PM To:     'Yacine El Mghazli'; rap@ops.ietf.org; Rawlins, Diana Cc:     Iliff, Tina; Joel M. Halpern; Brian E Carpenter Subject:        RE: [Diffserv] CountActTable

Speaking for my company, yes we feel it is very important to have such counters via COPS reports. Because SNMP is not a very scalable way to get counters on a large number of objects (e.g. IP interfaces which may be as numerous as tens of thousands or more on a given box) in a reliable way. I hope I will not get flamed for this statement!  ;-)

Plus differentiated services typically come with differentiated charging models (and some of these billing models might be based on volume). So it is key to get service usage (aka accounting) data in association with Diffserv policies. And in a reliable/scalable way.

This being said, based on the accounting framework currently defined in the "rap" group, accounting PIBs are clearly separate from the related "provisioning" PIB. And related objects (e.g. classifiers & meters) are referenced via pointers.

So it makes sense to not have counters in the Diffserv PIB. As long as there is a "Diffserv accounting PIB" in the works (or the general one described in the accounting framework is applicable). Jerome

=========================================== Jerome P. Moisand Senior Product Manager Unisphere Networks One Executive Drive  Chelmsford, MA 01824 Tel: 978-848-0648  Fax: 978-848-0399 mailto:jmoisand@UnisphereNetworks.com



-----Original Message----- From: Yacine El Mghazli [mailto:yacine.el_mghazli@ms.alcatel.fr] Sent: Thursday, April 05, 2001 4:57 AM To: rap@ops.ietf.org; Rawlins, Diana Cc: Iliff, Tina; Joel M. Halpern; Brian E Carpenter Subject: Re: [Diffserv] CountActTable

In the framework of COPS-PR configuration, is it important to be able to install/remove/report counters via the DiffServ PIB ? What do think people involved with the accounting PIB ? To my mind I hardly understand why one could not install a counter with COPS-PR in order to monitor it by SNMP, or even to report it via COPS-PR... I'm also interested in the reasons which made DS PIB authors remove the CountActTable from the PIB... I guess they must have good ones !

Thanks Yacine

----- Original Message ----- From: "Brian E Carpenter" <brian@hursley.ibm.com> To: "Iliff, Tina" <Tina.Iliff@WCOM.Com> Cc: <diffserv@ietf.org> Sent: Wednesday, April 04, 2001 9:25 PM Subject: Re: [Diffserv] CountActTable

> I don't see the issue for the MIB: it does have counters and nobody is > proposing to take them away. If you want counters in the PIB, now > is the time to say so since there will be a new version very soon. > The model isn't going to answer that question for you. > >    Brian > > > "Iliff, Tina" wrote: > > > > Here are some excerpts from diffserv-model-06: > > > > At the lowest level considered here, are individual functional datapath elements, each with their own configuration parameters > > and management counters and flags. > > > > using a Counter element downstream of the Meter, it might also be used to help in collecting data for out-of-band management > > functions such as billing applications. > > > > I find these excerpts unclear as to determining if a PIB or MIB or both should implement a data structure to support the > > definition of a Counter Action. > > > > Tina Iliff > > > >           -----Original Message----- > >           From:   Joel M. Halpern [mailto:joel@longsys.com] > >           Sent:   Wednesday, April 04, 2001 12:14 PM > >           To:     diffserv@ietf.org > >           Subject:        RE: [Diffserv] CountActTable > > > >           The Count Action is a diffserv packet handling action. Including it in a > >           configuration causes counting to be done.  It sure seems that this action > >           ought to be configurable with the PIB, just like every other Diffserv > >           packet handling action (Meter, Mark, Drop).  Otherwise, one would have to > >           use SNMP to CONFIGURE this element in order to later use SNMP to retreive > >           the counts. > > > >           This Action is not itself a counter that would go in an Accounting PIB, or > >           only in an SNMP MIB. > > > >           Yours, > >           Joel M. Halpern > > > >           At 03:13 PM 4/4/01 +0000, Iliff, Tina wrote: > > > >           >Yacine, > >           > > >           >Like I said, it is just an assumption or a guess.  Someone else will have > >           >to answer.  The Accounting PIB may be a good place for a > >           >counter.  However, I have not taken a look at that draft in a while. > >           > > >           >Tina Iliff > >           >-----Original Message-----  From:   Yacine El Mghazli > > >[<mailto:yacine.el_mghazli@ms.alcatel.fr>mailto:yacine.el_mghazli@ms.alcate > >           >l.fr]  Sent:   Wednesday, April 04, 2001 10:07 AM  To: Iliff, Tina; > >           >diffserv@ietf.org  Subject:        Re: [Diffserv] CountActTable > >           > > >           > > I am just guessing here but it makes more sense to me to only have a > >           > count  > action in the MIB and not in the PIB.  I do not consider > >           > counting of  packets  > traversing certain datapath functional elements a > >           > piece of policy;  however,  > I do consider it as part of a network nodes > >           > management. > >           > > >           >Do you mean that counting packets is only managed using SNMP ? > >           > > >           > > Tina Iliff  >  >  > -----Original Message-----  > From: Yacine El > >           > Mghazli  > > >           > [<mailto:yacine.el_mghazli@alcatel.fr>mailto:yacine.el_mghazli@alcatel.fr] > >            >   > Sent: Wednesday, April 04, 2001 7:39 AM  > To: diffserv@ietf.org  > > >           > Subject: [Diffserv] CountActTable  >  > In the lattest DS PIB :  >  > > >           > Quoted from the draft-ietf-diffserv-pib-03.txt (page 5) :  > > Action > >           > Tables  > >      A general extensible framework and examples of  > > >           > parameterization  > >      tables for Absolute Drop, Mark and Count > >           > actions.  >  > and page 11 :  > > 4.5.1.  DSCP Mark Action PRC  > > > >           > (...)  > > 4.5.2.  Absolute Drop Action  >  > So why is there no Count > >           > Action Table in the PIB ?  > However there is an CountActTable in the > >           > MIB...I feel  > confused.  >  > Thanx  >  > -- Yacine El Mghazli  >  > > >









    > ---------------------------------------------------------------------- > > >            >   Alcatel R&I  >   Software and Services Strategic Program - > >           > VIPeR  >   Marcoussis, France  >   Tel  +33 1 69 63 41 87  > Fax  +33 1 > >           > 69 63 11 69  >   yacine.el_mghazli@ms.alcatel.fr  >  > > >









    > ---------------------------------------------------------------------- > > > _______________________________________________ > diffserv mailing list > diffserv@ietf.org > http://www1.ietf.org/mailman/listinfo/diffserv > Archive: http://www2.ietf.org/mail-archive/working-groups/diffserv/current/maillist.h tml >