James, further to last night's critique:
I would suggest that the popular misuse of the lexicon surrounding the marketecture of this space has become so pervasive at this point that it transcends individual vendors' product description spins, and is now so widely out of control that we are giving full credence to their new meanings (the terminology used) whilst we attempt to rationalize what they actually mean and what's actually going on.
Another way of putting the same thing: Everyone is so hung up on escaping legacy
frameworks that we tend to entirely ignore, sometimes simply forget, the fact that we
are actually using those same legacy constructs as a means to escape them. Leveraging the investments in embedded infrastructure, and all that.
All DWDM gear along with the much celebrated terabit routing gear used for long haul inter-networking utilizes Layer 1 interfaces which are still defined entirely in SONET terms.
It would be farcical at this time to suggest that SONET and many other legacy characteristics have been replaced simply through the implementation of optical wavelength technologies. Some have, but there's a long way to go.
We thought that we were going to see optical routing. What we have, instead, up to this point in time, is a number of variations of multiprotocol label switching (MPLS) using OC-48 streams welded to SONET ports.
Whether we call this wavelength routing protocol (WaRP) from Monterey, or virtual transport routing protocol from someone else, it's still a Layer 2 technique much akin to ATM or MPLS. We've not yet moved very far from legacy routing and switching in this respect, we've just made it faster by switching lambdas instead of virtual ATM paths. And we certainly are not actually "routing" at optical speeds yet, we're barely able to reach the next level of traditional forwarding and lookups due to the increasing levels of software-intense policy decisions in increasingly complex networks.
We are simply lopping a lot more like traffic types onto the same lambdas through the use of SONET-defined containers, thus far.
The only DWDM and terabit box interfaces which are NOT SONET-defined at the present are those which are used in proprietary systems supporting METRO apps, such as Gigabit Ethernet (GbE), Fibre Channel (FC), IBM's mainframe channel enterprise system connectivity (ESCON), and perhaps some high-speed parallel interface (HIPPI) applications, somewhere. Some of the latter, too, are encapsulated in SONET containers, depending on the vendor involved. Otherwise, look for SONET defined pipes, lambda-ized or otherwise.
Sycamore, Juniper, Corvis, CIEN, NT, LU, Pirelli, they all abide by SONET
interfacing standards up to OC-192, with many of them now seriously flirting with
OC-768, and some already talking beyond. I don't see it going away anytime soon.
Perhaps there will be a greater uptake in GbE-over-lambda for ISP peering purposes
and by some enterprises at some point, but that remains to be seen.
It's the "containerization quality" of the SONET standard which has clung to the
industry, and used so extensively, because of its pervasiveness in add-drop muxes and
cross-connects which are still absolutely essential elements in reaching the vast majority
of end users (who do not have dark fiver installed). SONET has not survived due to its
other administrative baggage which it possesses in its upper layers, but it's still
SONET-framed, nonetheless.
A great deal of the "other baggage" is mitigated or removed by nulling out some of the
fields in the SONET overhead protocol, thus "skinnying" SONET down to the point
where it is almost invisible. But it's still there. One of the features which tends to get
skinnied out is the self-healing action, I should add.
Time flies. More later, Frank |
