A Last-Mile Solution to Broadband Access

3.4.05 -- According to Anthony Acampora, a professor of Electrical and Computer Engineering at UCSD and a member of the Center for Networked Systems (CNS), of the 900,000 commercial buildings in the U.S., 95% are within one mile of connection to fiber, but only 5% are penetrated by fiber.

 Anthony Acampora
Anthony Acampora

It’s the classic “last-mile problem”: The fiber can be accommodated by digging a ditch for it, but that’s prohibitively expensive. So how do you provide broadband access to 95% of the locations easily?

The solution, says Acampora, based on research he, Rene Cruz, also a professor in ECE, and their students presented recently at a CNS research review, is free-space optical communication.

This solution, though, doesn’t come without its own set of problems. The signal loses power over long distances. It diminishes when it has to transit rain, snow, and especially fog; in the latter case, the signal can suffer a loss of up to 50-300 dB per kilometer. It’s also necessary to keep the beams carefully focused between transmitter and receiver, and the two need to be within line of sight of each other.

“The result is that people have been considering – and rejecting – this approach for decades,” says Acampora. “But we think this conclusion is misguided.”

Perhaps the best approach in this situation, says Acampora, is a mesh topology. This approach makes it possible to keep the links sufficiently short to avoid, for example, an outage during fog. When distances are kept to less than 150 meters, for example, you can obtain the famous “five 9s” of availability: 99.999%. Distances can be extended by using buildings located near-in to the hub as regenerators of the signal to reach buildings further afield. This way a single point of fiber can reach a wider set of end points without the constraining requirement of a rectangular grid.

 Rene Cruz
Rene Cruz

In this context, Acampora and colleagues have developed a free-space optical mesh in software.

“You can quantify the capacity of the mesh,” says Acampora. “If C equals the number of buildings you want to serve, then 5C is the limit of the traffic. If a single building can be kept to generating a fraction of C, then the number of buildings that can be served increases.”

This project has led to exploration of a number of software issues. Given the locations of the buildings to be served and optimal locations for the base stations, how do you define and architect optimal connectivity? What routing do you use? What about managing mobility: What if, in addition to Internet connectivity, you want to support cell base stations? How do you manage faults and dropped signals transparently to minimize inconvenience to users?

Perhaps the most interesting question relates to complexity management: How do you derive rules for operating that network that represent an acceptable tradeoff that maximizes usable capacity yet simplifies software implementation?

The conclusion of the research team is that wireless optical meshes can provide low-entry, cost-effective, local broadband access at high capacity that can be delivered flexibly over a wide service area. The system is easy to install, and the equipment easily aligned. There’s no need to dig ditches or acquire spectrum license. High availability is maintained through short links that can be regenerated through optical links and mesh rerouting to achieve longer distances. Such a network can carry, simultaneously, voice, IP, video, and ATM cell streams. The network is easy to administer and manage, and problems can be diagnosed easily. And, when needed, wave-division multiplexing can increase capacity significantly.

“We think this is an attractive solution for any building that doesn’t have fiber directly to its doorstep,” says Acampora.

Related Links
Center for Networked Systems

Related Articles
Allocating Resources in a Federated, Distributed Computing Infrastructure
'Programming' the Net