ossy,
"I heard that it was too complicated to -technically- go beyond 2.5Gbit/s (STM-16 in SDH) was too complicated hence DWDM came along. What your views on that if you may?"
Sure. It's a complicated issue that is influenced by some not-so-apparent factors. To put some issues to rest, though, it wasn't long ago that OC-48 was being hailed as the end of the line, because higher rates were too complicated, not only as a function of feasibility to reach higher TDM rates, but for reasons that had to do with packet header recognition, buffering, table lookups, forwarding, etc.
This, as we now know, is no longer the case, and OC-192 ports will be handling all of the above in a growing number of vendors' products. And some have already predicted their ability to read and forward packets at 3072 "just around the corner." I don't know about that, but in time it will be feasible. These are all in electronic domain at the point of reading and recognition, by the way. The holy grail, however, would be to someday be able to read headers in the optical domain.
I note that George Gilder and others in the spotlight have dismissed higher TDM rates, in favor of lower orders of throughput based solely on lambda flows. There's a spectrum grid issue here, where we find that the "virtually infinite" nature of fiber is not all that infinite at all. It has to do with administration and the current limitations of static wdm models which employ "spectrum grids" which define the degree of bandwidth and separation that exist between lambdas in a wdm.
ITU 200 GHz and 100 GHz grid spacings, now down to 50 GHz in some vendors' wares, seem to dictate real estate parameters and how many individualized flows you can have on a given fiber. In theory, the lower the rate of each flow, the lower the amount of spectrum that you would need to hog up, per any given system design's parameters.
So, in this case where low bit rate-supported lambdas are prevalent, more lambdas are possible on a single fiber's optical grid. Very close spacings are possible, say at 1 or 2 GHz spacings or lower, as opposed to 200- or 50- GHz spacings.
It therefore behooves lambda-ites (as opposed to TDM-ites) to stay below a certain throughput threshold, so that more optical flows can proliferate on the same fiber. Whether this satisfies Internet backbone providers needs is another story.
These considerations in their most extreme form may be short-lived, however, as better- and more dynamic- optical filtering schemes evolve, and as improvements occur in tunable lasers and detectors allowing multiple line widths to be dynamically shared on a dynamically variable grid. But when a grid is dynamically variable, is it still a true grid? I guess so, if a number of lambdas or "channels" remain statically defined.
At present, it appears that the state of the art (with the possible exception of AVNX and a few others) supports a model whereby the optical grid is comprised of lambdas that are basically all of the same width.
One size fits all does not benefit the lambda-ite. They will prefer dynamic and variable adaptation schemes.
Of course, this doesn't really have anything to do with the inherent feasibiliy of reaching OC-3072 rates (160 Gb/s TDM), per se. However, these factors do play into the motivation that the industry, as a whole, demonstrates it has for getting to those higher TDM rates. Personally, I think that xSPs and other "networkers" will favor lambdaism, while backbone providers will look to higher TDM rates.
FAC |
