>At 12:10 PM 12/3/00, Howard C. Berkowitz wrote:
>
>>When we are talking about futures, the reality is that we truly
>>don't know. To say that carrier-scale backbones will be ATM
>>(probably not), POS, IP over raw DWDM, MPLS over raw DWDM, etc., is
>>not yet a given.
>
>Thanks for turning this into a logical, well-written discussion
>instead of a harangue, Howard.
>
>Probably not ATM for carrier-scale backbones? Is it because MPLS is
>looking like a better model? I realize I'm asking you to look into
>your crystal ball, and that's always hard, but I'm interested in the
>technical reasons off the cuff. (I know it could require a whole
>book to give a true answer! &;-)
Well, I just wrote one, although I don't think there is a firm
consensus as yet on what will win. Shameless plug: _WAN Survival
Guide_, ISBN 0-471-3428-3 (Wiley). Published just in time for the
Christmas rush. :-)
But yes, I believe MPLS is a superior model to ATM. Lots of people
call it "ATM without cells," removing the problem of the cell tax.
MPLS does not replace IP, but is an "overdrive" for IP. The biggest
confusion about MPLS, which I found in a number of Cisco
presentations, is that MPLS is far more than the label-switched paths
themselves. These paths need to be set up by routing mechanisms.
When I say routing mechanism, I include both dynamic IP routing
protocols and extensions for traffic engineering. RSVP-TE, LDP, and
CR-LDP don't replace routing, but distribute label switching
information that is based on routing (including traffic engineering
overrides to routing).
Many of the "optical routing" discussions really aren't talking about
something radically different, but, for example, identifying a path
by lambda rather than explicit label.
>
>Priscilla
>
>>We face challenges such as "is it better to have single 40 Gbps
>>OC-768 streams or multiple OC-192 over DWDM?" There are many
>>routing versus switching arguments, and MPLS is a mixture of the
>>two (even though there's intense religion about LDP, RSVP-TE, and
>>CR-LDP). We don't know the situations in which photonic switching
>>of lambdas is enough, versus photonic routing of individual
>>packets. Lots of things we don't know.
>>
_________________________________
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
