IP: RE: SEE THE LIGHT BOUNCE: Edupage, March 26, 2001

[David Farber <dave@farber.net>]

John,

The real question on the table is : is packet switching at all rational in 
the all optical networks of the future. My estimate is the answer is no. A 
long time ago Len Kleinrock and myself independently wrote articles 
strongly suggesting that the nature of very high speed networks was not 
friendly to TCP/IP and the nature of optics is not, in my opinion , 
friendly to conventional packet switching.

Dave

At 05:59 PM 3/26/2001 -0800, you wrote:
>From: John Shoch <shoch@alloyventures.com>
>To: "'farber@cis.upenn.edu'" <farber@cis.upenn.edu>
>
>Well, let's not get completely buried by the hype...
>
>--I hope everyone understands that the "current generation of switches"
>described here is (I presume) the electro-mechanical switches used to
>reconfigure optical links.  The "all optical" MEMS and bubble systems have
>switching times which make them useful for configuring and reconfiguring
>optical links, but they do not switch at rates which allow them to become
>real-time packet switches.  Thus, they will not displace real packet routers
>from Cisco, Juniper, et al.
>--In-line signal conversion may add expense, but it does not (by itself) add
>congestion.  Optical signals can be amplified, but over time, and distance,
>the signal degrades.  Optical signals passing through optical switches will,
>in many cases, eventually have to be re-generated -- which will still be
>done by going through an O-E-O (optical-electronic-optical) conversion.
>
>John Shoch
>Alloy Ventures




For archives see: http://www.interesting-people.org/