It sure is fun to try to understand these things with nothing available to define what they are.
We can consider, first, those nodes which serve as junction points [typically, digital cross-connect nodes and add-drop multiplexer points],
I don't believe that it is an add-drop multiplexer in the conventional sense. I believe the adds or drops are channel independent. That is, they aren't muxed onto a beat carrier of the general stream, rather they are muxed onto or off of a marked feed. The add equipment takes an RF input, assigns a sampling frequency at the base and mixes it. So the whole lambda bandwidth isn't used to receive the input, only the designated channel residing in a discrete portion of the lambda's spectrum. Each component channel carries a unique designation so it is easy to tune your splitter to separate off which component you want at the receive point. How does receive know which signature has been assigned? They email it!
where larger flows are commonly groomed and filled with voice grade DS0s, T1s/T3s, and where exchanges and handoffs take place between
disparate carriers or ISPs, or even enterprises and their service providers.
Grooming, filling, and disparity is minor. That's all upper level stuff. You don't get into someone else's soup. You just take whatever protocol or form, dump it in a channel, and send it on its way. No need to transform, only a need as with an ATM header to send it to the right place so that they can worry about de-grooming, de-filling, de-sparity, etc. Are we on the same page?
I will assume that if an OC-48 or OC-192 optical stream is carried over one of SR's systems, then that OC-x signal will at some point need to be extracted from their larger flow, using a tuned RF method of some sort.
Yes, a splitter. This is an optical mechanical device whose moving parts move at the speed of light!
How is this OC-x broken down into its constituent T1s and T3s for distribution purposes, once it hits the remote SR device?
Didn't we go through all this many months ago? Ah, there's the problem. The company out sits its critics.
You take the OC-x and slap it on. You don't try to decompose it. You mark it for the junk it carries and send it to some other fool to decipher with the traditional junk the established mfgers make. Sr is not trying to redo the entire system. They are just trying to fit in on top of everyone's junk.
How will SR do this without some form of industry-compatible time slot interchanger in tandem with an array of ports capable of satisfying all of these feeds?
That would be swallowing their soup after swimming in it. The idea I think is to use something like Fedex and ship the whole pile without trying to disentangle all the knots. You can't sell something good, you have to fit with something that is bad. They call it evolution.
The answer to this is that it will need to enter a DCS, or
digital cross-connect system while preserving the integrity of
the OC-48 or the OC-192 at some optical level of the types
made by ALA or TLAB, etc. This may be difficult to do if
they ignore the conventions and the underlying syntax
contained in SONET headers.
QED.
For these reasons, two things appear to be likely to me at this time.
(1) that SR engines will be used primarily as "dense route"
aggregation vehicles, such as those required between core
routers on the Internet's backbones, and not those which
are needed between clusters of largely heterogeneous
traffic types (no mixed marriages, in other words); and,
(2) in the future, they will need to devise some method of
creating (accepting and launching) the same types of
channelized bundles which are now common to the rest of
the industry, from at least the T3 level on up, if not the T1
level.
Right. The only proviso is in (2). Somewhere above T3 you run into channel or lambda limit. Certainly any T-x single feed can be put on the wire where x < a multiple of maxbandwidth/channels.
Right now they are taking information from various sources and compactifying individual large flows (which are already aggregations of many smaller flows) onto an even denser flow. They must also be able to selectively de-constitute that flow at the usual nexus points and redirect subordinate streams (i.e., virtual tributaries) in ways which are similar to the mapping techniques used today by conventional central office equipment. [I know, I know, don't say it.]
The selection is easy as I mentioned above. It's the creation that is what the SR transceiver is all about. You have to embed, mark, and mix, at extreme clock speeds. It's hard to get them all jacked up and cooperating. Easy to get them down.
Otherwise, they will find themselves stranded in some imporatant ways, unless those who purchase their wares are accommodating in principle, and make the necessary adjustments to deal with such inflexible flows until the model has had some time to become more commonplace.
No accommodation in principle is necessary. The SR technique is brute force. You ship the whole mess and let the separate disentanglers which already exist do the dirty work they already are doing. The advantage is that this is simple and cheap in comparison to pumping up the existing crazy quilt crisscross disconnected land lines. Told you they out sat us. |
