An interesting little blurb on the importance of WDM/Optoelectronics:
Where Will The Bandwidth Come From?
Paul Green, IBM
I hear voices
Much has been made of the fact that the precipitous emergence of the Web occurred without any predictive planning, strategizing, corporate resource redirection,
targeting of research, or any form of top-down push, but instead seems to have occurred by the pull of societal needs of the user community, enabled (but not
driven) by technical innovation. From the communications technology standpoint, there seem to be several messages that these events are trying to send to us and
which we can try to decode and be guided by in order for the fruits of our researches to do the most good. The following is one (admittedly narrow) view:
Down with television
Many think that much of the appeal of the Web is its peer-peer nature (much like the traditional POTS) to a citizenry that is subliminally fed up with fifty years of
broadcast TV as the only wideband communication medium available. There is implied here an uneasiness and even resentment about having all the ideas being
generated by a group of arbiters of public opinion and thought whose small number is incompatible with democratic principles. When anyone can set up his/her own
home page and be a source of opinion, thought, commercial enterpreneurship, or whatever, then that's real empowerment. Today's ruling on the Communications
Obscenity question by the three-judge court seems to have caught this point accurately.
Location independence
The Web can be considered a sort of "third wave" in the democratization of computing - MIPs to the Masses. The first was the PC, but it had the clumsy DOS
command line interface. The second was the MAC with the point/click GUI, but the objects and processes were all local, no farther away than the milliseconds of
hard disk access. With the third, the Web, the objects or software didn't have to be local but can be anywhere. (One could conjecture a future fourth wave in which
the accessor of these resources could also be anywhere).
Bandwidth pigs
In order to keep the access time down, not only is it necessary for routers and servers to have faster throughput, but the communication links need to be speeded up
too. Moreover, every new Web innovation (voice, video, JAVA downloads) seems to exacerbate the situation. It has been widely predicted that the year 2000 will
be the year that data traffic will finally exceed voice traffic in the telco backbone and that after that, the traditional factor of roughly 1.5 per year in bandwidth per
user will suddenly steepen to the present factor of eight per year bandwidth per user seen in the Internet since the Web came aboard. Already, the largest Internet
backbone carrier (MCI) is replaing his DS-3 (45 Mb/s) Network Access Points with OC-3 (155 Mb/s) and reports that they now realize that within one year after
that they will have to do it again with OC-12 (622 Mb/s). To paraphrase Will Rogers, nobody ever went broke overestimating human greed for bandwidth.
Send in the fibers
If one looks ahead, it seems clear that the only technology capable of handling such growth rates in a reasonably future-proof way is fiber. Fortunately, over 10
million strand miles of it is installed already in the U.S. alone (and growing at a rate of 4000 strand miles per day). Equally fortunately, there are a potential 25,000
GHz. of capacity in each fiber (5,000 with erbium amplifiers, still an enormous resource), and people are learning to build WDM links and networks to exploit it.
Just as fiber replaced copper in a center-outward pattern (first the large central offices, then the small ones, then to the neighborhood, eventually to the desktop)
most of the WDM that is being installed is following the same pattern. Some fiber is already available to large companies, as my employer is already finding in selling
optical networking to business.
Transparency
Going back a few years, the reason we technologists have been able to build the present rich communication world on such a modest base as twisted pair copper,
with its paltry 3.5 kHz of bandwith, has not only been cleverness with coding, equalization and modulation, but the fact that the medium was transparent. Once the
connection is in place, it doesn't care about bit rate, framing or any higher level protocol parameters. WDM circuit switching offers the same sort of transparent,
futureproof base, but now with 8 orders of bandwidth and error rate improvement.
Research
Unfortunately for traditional communication research, this very transparency and seemingly unlimited bandwidth obviates the need for many traditional disciplines so
necessary and so fruitful for the limited-bandwidth noisy twisted pair world. Even congestion control largely goes away as an issue when communication is via a
"lightpath" in which a given wavelength over a zigzag path through the network acts just like a fiber of on's own. The research issues whose lack of solution is holding
back the unlocking of the fiber's bandwidth potential are not the usual ones, but development of clever ways of cutting optoelectronic component costs. There is
much that is not understood at the system architecture level, but this is still not the bottleneck that hardware cost is.
ana-www.lcs.mit.edu |
