The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2000-Jun> msg00185



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

"This is a gripe" [was Re: [Diffserv] MPLS Diffserv Extensions related questions/comments

  • From: "Manfredi, Albert E" <Albert.Manfredi@PHL.Boeing.com>
  • Date: Wed, 7 Jun 2000 18:20:31 -0400
  • Cc: diffserv@ietf.org, mpls@UU.NET

> -----Original Message-----
> From: Kimmo.Raatikainen@nokia.com [mailto:Kimmo.Raatikainen@nokia.com]

[ ... ]

> Please keep in mind that gigabit cards are already available. In our
> experiments we have found out that a high-end PC running 
> Linux can send at a
> rate of 250-300 megabits per second using TCP/UDP.

Actually, too bad Bora's "gripe" wasn't fully discussed on here. I guess it
didn't fit in with the router model work being done at the moment.

I'd like to know how one can offer "service differentiation" over a
packet-switched internet (small i) if one doesn't allow for network nodes to
examine header fields, and then make queuing decisions based on some
undefined and growing list of criteria? And potentially different criteria
for different networks, as far as that goes? 

Maybe some sort of more hardware-friendly scheme, where queuing decisions
are based on a simple set if rules associated with the 64 code points?
Dunno. But it would definitely lead discussions down a different path, no?

Bert
albert.e.manfredi@boeing.com