Re: [RRG] Tunnel fragmentation/reassembly for RRG map-and-encaps architectures

[Robert Raszuk <raszuk@juniper.net>]

Hi Tony,

> > That is good observation ... but I want to also point out that there 
> > is already a deployed technology to offer prefix independent convergence.
> >
> > So far deployed within an AS scope .. but very possible to be deployed 
> > system-wide if we go to two tier BGP hierarchy.
>
> Thanks for pointing that out.  I'm no expert in PIC, but it would seem 
> to address purely the cases where there is little state flux in the 
> control plane.  Consider the case of when a new router first joins the 
> BGP mesh today.  This is when we will see the biggest impact in the 
> number of prefixes carried.

That's as you may perhaps noticed I am a big supporter of two stage 
tunneling within each AS or within each POP. So only very limited number 
of routers per domain would need to go via this biggest impact cycle .. 
clearly not all of them. And as we all know we do have tools to allow 
for graceful initial convergence.

> > We have said that control plane is pretty much not the issue. So 
> > perhaps we should shift the discussion into ways to fit control plane 
> > into data plane so PIC would be achievable system-wide ?
>
> I guess I'm disagreeing with you.  The control plane is very much the 
> issue.  The control plane *mechanism* is not the issue, but the 
> scalability of the data manipulated by the control plane very much is at 
> issue.

I am not questioning that at all ... especially thinking about the scale 
involved going into the future. I am just trying to determine if current 
mechanics of the control plane .. when decoupled from being responsible 
for fast locator/next hop reachability vehicle could not meet the new 
scale requirements. And if not ... why.

I guess the real question get's down to either carry or not to carry all 
EIDs at each let's call it default mapper or even as in NERD each ITR. I 
 am not sure if on demand caching will be the answer to this question. 
Personally I doubt it.

Cheers,
R.


--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg