Optical Fiber Capacity Growth

kastaverious writes: "I found this on Scientific American. It talks about developments in all optical switching and the growth in capacity of optical fiber. The article has some interesting graphs of bandwidth demand and the growth in bandwidth availabilty. There is also a good explanation of some of the technical issues involved in increasing switching capacity, and efforts underway to overcome these problems." The article also has lots of good SciAm-style graphics. This short article at Janes also sheds some light on the world on undersea cable laying, which also recalls the article Neal Stephenson wrote for Wired a few years ago.

Comments

Articles accurate but too fluffy. Smoke & Mirrors

[Devout Capitalist]

The article misses the key technologies of the future summarized as the smoke and mirrors story I came up with over at Sun Microsystems. We used this to talk about honking bandwidth, the need for big servers, and why a portable language like Java makes sense.

DWDM is a start, but there are two major problems:

• Smoke:Right now we can throw a lot of bandwidth across a long haul fiber, but these use expensive lasers that run only one or two protocols. There are a lot of seperate networks, HTTP/TCP/IP, SONET, some voice stacks, even Telex. Each of these networks has its own protocol stack right down to some fiber based ethernet standard or hacked up 1990's protocol. The best solution is to make a 'smoke' box that will allow splitting by frequency so that I can run a dozen frequencies as SONET, a dozen as voice, and twenty TCP/IP. The magic 'smoke' box splits the incomming fiber into several seperate fibers, each carrying a distinct set of frequencies that can leverage other equipment. By combining together different inputs, I can use a single long fiber for multiple networks. One order of magnitude.
• Mirrors:This is the area where Lucent is making some progress. I need to do some nifty tricks with routing or my gross bandwidth buries my useful bandwidth. All the ATM switch cloth with IP cache in the world won't help if I need to cross over the optical/electrical boundry for every packet. A 'mirror' could be the simple stuff with Lucent using a physical switch to reimplement timesharing (1 cycle for SF to NY, 1 for SF to Boston, 1 for San Jose to NY, ...). The mythical mirror solution is to hit a lattice with a signal such that the reflection property reflects to a different destination. You would only need to cross the boundry for the destination part of a packet or routable stream. This 'mirror' magic would be an independent improvement from DWDM or 'smoke'. Most likely, you would use DWDM, split to fibers with smoke, route with mirrors. Another order of magnitude; maybe two.

Finally, give up on rewiring the last mile. The DSP and other signal processing tricks will get faster and cheaper more quickly than any solution that requires rewiring. It makes financial sense to swap end point electronics rather than rip open walls. You may see many more COs making shorter runs to the houses, but either existing coax or twisted pair into the house will carry our future bandwidth. (Thanks to Brent and Richard for convincing me.)

I miss Sun, they had more interesting problems than running a non-profit. See the non-profit at TrueGift Donations.

Cheers!

Charles