Re: [netext] Needs of traffic spec. on MAG?

["Laganier, Julien" <[e-mail]>]

Rajeev Koodli wrote:
>  
> I see this is a terminology issue.

I do not think so...
 
> What is considered unique prefix in the draft is that each access is
> free to assign whatever prefix is deemed right at the time of attachment

If a new prefix is assigned by an access, the LMA is aware of it (either new prefix is signaled by MAG in PBU sent to LMA, or new prefix is assigned by LMA itself.) That prefix is subsequently added to the BCE for the mobile node.

The new prefix is not tied to a specific access but is valid on any access, thus making it possible to move flows across access irrespective of the prefix to which they belongs. 

> - as folks have pointed out, this is already supported in 5213. Now, *whenever*
> the LMA decides to move one or more flows from a prefix for that access to
> another access, it would have to signal the corresponding MAG. Looks like we
> agree here.

I disagree. Since all prefixes assigned to a mobile node while attaching to whatever accesses belongs to that mobile node, a MAG only needs to be made aware of the complete set of prefixes to be able to perform its function -- forward packets between the mobile node and the network.
 
> What is considered a shared prefix in the draft is that the LMA assigns
> the same prefix(es) to the new attachment. In this case, the LMA does not
> need to signal the MAG for flow mobility purposes. However, we need to make
> some extensions to RFC 5213 signaling requesting LMA to assign the same
> prefix(es).
> 
> I hope this clarifies.
> 
> We need to support both models, regardless of what terminology you use.

I think that to support flow mobility the only model that we need to support is one in which all prefixes assigned to a mobile node can be used on any access to which the mobile attaches.

That is orthogonal to the function of a new access assigning an additional prefix that you hint at.

--julien

> On 8/2/10 11:51 PM, "Tran Minh Trung" <[e-mail]> wrote:
> 
> > You are right. After LMA deciding to move the flows, both MAG2 and
> > MAG1 will advertise p3. It means that the prefix p3 is now shared
> > across 2 physical interfaces.
> > That is the reason why, IMHO, we do not need to consider two
> different
> > scenarios, just consider shared-prefix model is enough.
> >
> > If we have a solution for the LMA to assign the same prefix(es) to
> > MAGs for the logical interface, then it is very easy to support flow
> > mobility.
> >
> > Regards,
> > TrungTM
> >
> >
> >
> > On Tue, Aug 3, 2010 at 2:22 PM, Koodli, Rajeev <[e-mail]>
> wrote:
> >>
> >> Even though the logical interface can make the packets available to
> the MN,
> >> the main issue is whether MAG1 forwards those packets.
> >> MAG1 needs to advertise p3 as well. It is for such cases we need
> signaling
> >> from LMA *whenever* the LMA decides to move the flow(s).
> >>
> >> -Rajeev
> >>
> >>
> >> -----Original Message-----
> >> From: [e-mail] on behalf of Tran Minh Trung
> >> Sent: Mon 8/2/2010 7:48 PM
> >> To: Koodli, Rajeev
> >> Cc: Laganier, Julien; [e-mail]; Youn-Hee Han; [e-mail]
> >> Subject: Re: [netext] Needs of traffic spec. on MAG?
> >>
> >> I agree that the prefix p3, according to RFC 5213 , does not have to
> >> be simultaneously assigned to MAG1. However when we use logical
> >> interface at the MN, the prefix p3 should be shareable to allow flow
> >> mobility.  In addition, the logical interface can receive packets
> sent
> >> to any of its sub-interfaces as described in
> >> draft-melia-netext-logical-interface-support-01 (property #2).
> >>
> >> The question here is how to change RFC 5213 to support shared-prefix
> model?
> >>
> >> Regards,
> >> TrungTM
> >>
> >>
> >> On Fri, Jul 30, 2010 at 7:45 AM, Koodli, Rajeev <[e-mail]>
> wrote:
> >>>
> >>> Single prefix on multiple MAGs is clearly one choice.
> >>>
> >>> We agree that the prefix has to be valid on an interface for the
> >>> corresponding flow to traverse the MAG.
> >>> If MAG1 has prefix p1 and MAG2 has p2, then the simplest form is p1
> and p2
> >>> are valid on both MAG1 and MAG2.
> >>> However, I should also be able to assign p3 to MAG2 (for example),
> which you
> >>> can do today with RFC 5213.
> >>> The prefix p3 does not have to be simultaneously assigned to MAG1.
> This
> >>> should be continued to be allowed with the flow mobility support.
> >>>
> >>> -Rajeev
> >>>
> >>>
> >>
> >>
> >>>
> >>> -----Original Message-----
> >>> From: [e-mail] on behalf of Laganier, Julien
> >>> Sent: Thu 7/29/2010 4:49 AM
> >>> To: [e-mail]; Youn-Hee Han
> >>> Cc: [e-mail]
> >>> Subject: Re: [netext] Needs of traffic spec. on MAG?
> >>>
> >>> Carlos Jesús Bernardos Cano wrote:
> >>>>
> >>>> Dear Youn-Hee,
> >>>>
> >>>> On Wed, 2010-07-28 at 21:43 +0900, Youn-Hee Han wrote:
> >>>>> Dear Bernardos,
> >>>>>
> >>>>>
> >>>>>
> >>>>> I'd like to ask a question about
> >>>>> "draft-bernardos-netext-pmipv6-flowmob-00."
> >>>>>
> >>>>> In the call procedure figure, I noticed that traffic
> specification is
> >>>>> delivered to MAG from LMA.
> >>>>>
> >>>>> I understand that the specification should be managed at LMA to
> do a
> >>>>> traffic filtering.
> >>>>>
> >>>>>
> >>>>>
> >>>>> However, is such a traffic filtering still needed at MAG?
> >>>>
> >>>> The MAG needs to know the flow that is gonna be routed through it,
> so
> >>>> it can insert the required routing state.
> >>>
> >>> If a single prefix is use across all of the MN accesses then
> there's no need
> >>> for this.
> >>>
> >>>>> IMHO, it is not needed at MAG. If so, why is the specification
> >>>>> delivered to MAG?
> >>>>
> >>>> This is something that can be discussed. The minimum info required
> is
> >>>> the MN's prefix of the moved flow, so the route is installed at
> the MAG.
> >>>> If finer control is required (e.g., prevent the MN send flows
> through a
> >>>> MAG that is not the one the LMA has setup flow mobility state, or
> other
> >>>> purposes), then the flow mobility (traffic selector, e.g., 5-
> tuple)
> >>>> should be delivered to the MAG.
> >>>
> >>> I don't think this is needed.
> >>>
> >>>>> Isn't sufficient that the home network prefix related to the
> traffic
> >>>>> is delivered to MAG?
> >>>>
> >>>> That'd be the minimum info. As mentioned before, the MAG may need
> the
> >>>> whole flow definition, and in that case the LMA has to deliver
> that
> >>>> information to the MAG.
> >>>
> >>> If a single prefix is use across all of the MN accesses then
> there's no need
> >>> to do this.
> >>>
> >>> --julien
> >>> _______________________________________________
> >>> netext mailing list
> >>> [e-mail]
> >>> https://www.ietf.org/mailman/listinfo/netext
> >>>
> >>> _______________________________________________
> >>> netext mailing list
> >>> [e-mail]
> >>> https://www.ietf.org/mailman/listinfo/netext
> >>>
> >>
> >>
> >>
> >> --
> >> Ph.D., Senior Member
> >> Electronics and Telecommunications Research Institute
> >> Standards Research Center
> >> 161 Gajeong-Dong, Yuseong-Gu, Daejeon, 305-350, KOREA
> >> Tel : [phone],   Fax : [phone]
> >>
> >> _______________________________________________
> >> netext mailing list
> >> [e-mail]
> >> https://www.ietf.org/mailman/listinfo/netext
> >>
> >
> >