You asked what was at issue with ethernet or why EPON is far away or why ethernet developers aren't 6 months from extinction or why SDM in the ECI might be leveraged into big or why standards building can hurt more than help?
From Riverstone's site:
Importing the LAN to the MAN - in particular, using Gigabit and 10 Gigabit Ethernet technology - offers another attractive approach to revitalizing the MAN. While Ethernet technology is certainly not a "pure" optical networking approach, it does represent a very cost-effective and practical means of bringing big bandwidth to the MAN. And to a greater extent than the save-SONET approach, Ethernet MAN's natural comfort with IP-traffic suggests healthy long-term prospects.
The approach is deceptively simple. In recent years, the Ethernet - originally a broadcast standard - has developed full duplex, switched manifestations: Gigabit Ethernet, and 10 Gigabit Ethernet. With these developments, Ethernet technology has become suitable for point-to-point networking, as in the MAN or WAN. From there, it becomes simply a matter of sending Ethernet frames over dark fiber; that is, specifically not relying on a particular physical transport standard, like SONET.
The advantages of bringing Ethernet to the MAN can be stated clearly. First, Ethernet is a natural fit for IP traffic. Ethernet and IP packets come in similar sizes, meaning very little waste. Ethernet, moreover, is such a popular LAN standard that it is likely to be found at both ends of a connection - and of course, the fewer frame conversions the better. But the principal advantage of Ethernet is simple - more bandwidth for much less money with a trusted, simple technology underlying the enterprise. The motto of Ethernet's backers? "Twice the bandwidth at half the cost."
Ethernet does, however, face its own problems stemming from the Ethernet's LAN origins. First, Ethernet MAN solutions do not leverage the existing SONET infrastructure, unlike the Save-SONET solutions, and to an extent, WDM. Second, Ethernet does not, by itself, have the kind of advanced grooming and QoS provisioning functions that other, more complex protocols provide.
There is an important answer to this last point - Multiprotocol Label Switching. MPLS allows (among other things) an Ethernet network to enjoy the kind of QoS provisioning usually native to connection-oriented protocols like ATM. And if that were not enough, some of Ethernet's backers would suggest a second answer: Cheap bandwidth will conquer all. The ultimate question is whether a technology lacking the "natural" advanced provisioning features of connection-oriented approach like ATM-SONET can deliver in the more rigorous MAN environment, while at the same time overcoming the infrastructure "head-start" of the save-SONET movement. But simple, best-effort standards have a long history of beating more complex full-feature protocols - this is the story of TCP/IP. 10 Gigabit Ethernet, for all these reasons, is the one to watch. |
