Re: [DNA] DNAv6 Review from Mobopts

[JinHyeock Choi <jinchoe@gmail.com>]

Dear Hemant

Thanks for your kind elaboration, which made me understand the cable
case the better, I hope. :-) I assume that, in CMTS, 'bundle of L2
links' form a kind of single (virtual) L3 link. While working on DNA,
I found out that the concept of 'link' is more complicated than I had
expected. :-(

Another thanks for your trouble and it would be nice if you comment on
Simple DNA draft.

best regards

JinHyeock


On Jan 11, 2008 2:40 PM, Hemant Singh (shemant) <shemant@cisco.com> wrote:
> JinHyeock,
>
> My humble apologies for jumping in late on this thread. Let me explain
> why the cable deployment is good and not as messed up as one might think
> - I will explain the details. I don't consider the cable network as a
> multi-link network. It's more of a point to point network. Here is why.
>
> The cable router, which is also called a CMTS (Cable Modem Termination
> System), has a property of a virtual bundle network interface. CMTS is
> also a DHCP relay agent. The L3 virtual network interface aggregates a
> bunch of physical L2 cable interfaces. For example, a Cisco CMTS router
> can have 40 cable network L2 interfaces - each of the 40 L2 cable
> interface can serve about 5000 homes. It doesn't make sense for the
> cable operator to assign L3 IP addresses to all of the 40 cable network
> interfaces - it's a configuration hassle for operators. That is where
> the L3 virtual bundle interface comes into play. One can aggregate say,
> 20 cable interfaces, under one bundle interface on the CMTS router. So,
> now the CMTS with 40 cable network L2 interfaces has only two L3 virtual
> bundle interfaces to which IP addresses and subnets are assigned. For
> IPv6, every bundle interface on the CMTS is one link. Yes, the multiple
> L2 cable interfaces represent physical links going to various homes, but
> the single L3 virtual bundle interface serves all the homes.
>
> Further, the physical connectivity of the cable deployment is such that
> no host behind a cable modem can communicate directly to another host
> behind another modem. All host traffic has to go through the CMTS
> router. Since hosts send traffic directly to router to communicate with
> any other host, that is why cable deployment if always off-link for both
> IPv4 or IPv6. That is also why the cable model maps to a point to point
> model.
>
> That is why cable folks were looking at ND to see how to specify
> off-link in ND for CMTS router. After going thru the ND RFC's we
> realized there was ONLY one way one could signal off-link in ND. That
> was by not sending any PIO (Prefix Information Option) in RA.
>
> When I get the time, I will review DNA simple draft as well and send my
> comments.
>
> Kind Regards.
>
> Hemant
>
> -----Original Message-----
> From: JinHyeock Choi [mailto:jinchoe@gmail.com]
> Sent: Thursday, January 10, 2008 6:54 PM
> To: Wes Beebee (wbeebee)
> Cc: Bernard Aboba; dna@eng.monash.edu.au; Hemant Singh (shemant)
> Subject: Re: [DNA] DNAv6 Review from Mobopts
>
>
> > >  thanks for your sharing ideas with us. I wish you would also
> > > comment
> > on the existing simple DNA solution to facilitate its advancement.
> >
> > Sure thing.  Please send me a copy of the current drafts/point me to
> > their location and I will look at it when I have the time.
>
> http://www.ietf.org/internet-drafts/draft-krishnan-dna-simple-01.txt
>
> best regards
>
> JinHyeock
>