The Routing Over Large Clouds Mailing List Archive by date

[Curtis Villamizar <curtis@ans.net>]

In message <rgc.1150500317A@hogpa.ho.att.com>, "Robert G. Cole" writes:
> Dave,
> 
> Re: the NHRP draft, it reads:
> 
> >   If the NHRP request is triggered by a data packet, station S may
> >   choose to dispose of the data packet while awaiting an NHRP reply in
> >  one of the following ways:
> >
> >     (a)  Drop the packet
> >     (b)  Retain the packet until the reply arrives and a more optimal
> >          path is available
> >     (c)  Forward the packet along the routed path toward D
> >
> >   The choice of which of the above to perform is a local policy matter,
> >   though option (c) is the recommended default, since it may allow data
> >   to flow to the destination while the NBMA address is being resolved.
> >   Note that an NHRP request for a given destination MUST NOT be
> >  triggered on every packet, though periodically retrying a request is
> >   permitted.
> 
> If the normal routed path over the NBMA travels thru routers A, B, C and D,
> where routers B and C are transit routers,  wouldn't this lead to
> the generation of three seperate "short cut" link layer connections (in the
> case of a connection-oriented NBMA like ATM) for the same packet?
> For example, 
>     - router A receives a packet and determines that it is to be forwarded
>         to its NBMA interface.  it sends a NHRP request and forwards
>         the packet to router B,
>     - router B receives a packet and determines that it is to be forwarded
>         to its NBMA interface.  it sends a NHRP request and forwards
>         the packet to router C, etc
>     - once the NHRP replies return to routers A, B and C, they each establish
>         their own "short cut" connections to D.
> 
> If this is indeed the case, then option (c) above should be removed in favor
> of one of the remaining two.  Am I missing out on something?
> 
> Thanks,
> 
> Bob Cole
> 
> 
> Robert G. Cole
> AT&T Business Multimedia Services, Technical Marketing
> rgc@qsun.att.com              +1 908 949 1950 (voice)
> attmail!rgcole                +1 908 949 8887 (fax)
> 
> AT&T Bell Laboratories
> Room 3L-533
> 101 Crawfords Corner Road
> Holmdel, NJ  07733-3030
> USA


Bob,

Major transit routers today are seeing on the order of 10,000 unique
destination addresses over short periods of time (90 second sampling).
If a transit router does not wish to support 10,000 VC (keeping in
mind that the entire VCI space is 64K), then it would probably be
configured to forward packets hop by hop only.

I don't think deleting option (c) is needed.  This is a usage issue
more than anything.

Curtis


ps- In data I took a year ago we saw 7,800 dests at 6 kpps over 90
seconds, if I remeber correctly.  This was single interface on one
router.  Only a few samples were taken, at different routers, and the
range was about 4,000 to 7,800.  This is all from memory, but I've
been asked about this enough times that I'm fairly sure the 7,800
figure is accurate.  Traffic has more than doubled since then.