The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2000-Oct> msg00258



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Traffic engineering and RSVP

  • From: David Charlap <david.charlap@marconi.com>
  • Date: Thu, 19 Oct 2000 17:35:35 -0400

Rajeev Manur wrote:
> 
> I never said anybody would or would not do 5 tuple lkup in the core.
> All i am saying is that your statement "Hence data path becomes slow"
> does not make any sense, because as far as i know today's packet
> processors can handle this without affecting the line-rate
> performance..

What do you define as "line rate"?

OC-3?  OC-12?  OC-48?  OC-192?  Faster?

And how many ports per switching fabric?  4? 16? 256? more?

No matter how fast the chips in your switching/routing processor can do
5-tuple lookups, there are environments where they are not fast enough.
And these environments are what you typically find in the core of major
networks.

Customers are looking for ever-increasing line rates, and increasing
port densities.  And their demands are increasing faster than the
ability of chip vendors to implement IP routing-table lookups. 
Especially if those lookups involve more than just matching destination
addresses.

-- David