My wonder is whether these hybrids--still in the electronic domain--will give way to an optical interface between the optical networks and the IP. Of course, that would require "smart" light--which we don't have yet.
Discontinuities are killers. According to the Gilder "lambdasphere" routers at the core may someday become a thing of the past. In fact this vision has already been proven possible in the laboratory. In such a network every connection is a point to point connection. The entire network reverts back to the simpler time of the small town switchboard for phone service (picture operator moving plug to plug for a direct connection to each premise. Of course this is done much more elagantly and automatically in the all optical network, but the analogy is apt).
Cogent is in the process of building an "all-optical" network and talk about a discontinuous product offering:
cogentco.com
cogentco.com
100 Mbps at $1000 a month vs. existing T1 1 Mbps offerings at $1000 a month. Cogent may be the head bowling pin that topples the rest if it succeeds. No one can match this sort of pricing and servicing offering.
This network however is utilizing Cisco's layer 3 terabit core routers and only working at OC-192 speed in the long-haul and OC-48 in the metro. It is not avoiding the electrical to optical switching of this all optical network paradigm.
However, characteristic of the futur eall optical network, each 100 Mbps connection is exclusive to the user with Cogent using 128 channels per fiber.
Given that each companies' connection is exclusive, at least until it his the user's premise (where the enterprise's own infrastructure will take over), this would seem to set up the eventual transition to an all optical network in the future, where each user's site is identified by a unique light frequency and exclusive to that user. Cao estimates that 250,000 wave lengths should be sufficient for a network large enough to support 20 million users. Cao, like Moore before him, is predicting that this sort of capacity is inevitable, and Avanex technology is the key (we shall see: gorilla report pending).
But still not there yet. My guess is, is that things like optical scalable routers by the likes of Avici and certain to be created competitive offers by Cisco and Juniper will delay this transformation for some time as telecoms have turned out to be rather conservative in their architectural choices and would prefer to continue improving the current architectures. At some point they will reach their limit, but that point is not today. Networks are not in enough pain yet to move to the simpler, lamdasphere concept even if all the necessary technological parts were available today. If Cisco et al., have their way, networks never will reach that threshold.
But if (or when) it happens, Cogent looks like a place it might happen first. In addition Sprint and Williams are also working on the idea (at least in the lab, if not in the field).
Tinker
Quick note: Xerox, of all companies, has just announced development of an all optical switch that they claim to have achieved 100% manufacturing yields using MEMS technology. This all optical switch seems to be a leap ahead of existing technology in competing laboratories. Perhaps like SDLI, Xerox might spin this one off.
But the reason Gilder has not listed any optical switching companies on his recommended list, is that in the ultimate Lambdasphere, which may start becoming possible quicker than we imagined (maybe 2-5 years) optical switching as we envision today from ONIS and Sycamore and Cisco is not really necessary (except in small quantities at the edges). Yeah, but Gilder's been predicting that one for quite some time. Have to continue to read the tea leaves and follow the value chains. The all optical network is discontinuous, no doubt about it. Be interesting to see how it evolves. |
