Optical network test beds blooming
- By David Perera
- Sep 24, 2004
A combination of available fiber-optic cable, government funding, promising technology and increasing awareness of widespread uses for highperformance computing has resulted in the growth of all-optical network test beds during the past year.
With real-time applications growing exponentially and outstripping the data capacities of today's networks, the need for light-wave networks extending to end users within the next three years is a practical mandate, said Bill Wing, a network
architect at the Oak Ridge National Laboratory's Networking Research Group.
"The technology and the need and
the ability to pay for it are coming
together fortuitously, and that fortuitousness makes it a prime target for
research," said Grant Miller, a technical staff member at the National Coordination Office for Information Technology Research and Development.
And with that has come an increasing realization that "if you've solved the problem locally, you haven't solved it at all," said David Nelson, the office's director. As a result, scientists met in California in August to hammer out network transfer protocols and solutions to common problems.
"It's not feasible for us to continue laying fiber in the ground for specific projects," said Kevin Thompson, the National Science Foundation's National Middleware Initiative program director. "We have maybe 20 different entities that are starting from scratch."
Because the testing sites are being built simultaneously, close collaboration will prevent researchers from "reinventing the wheel 20 different times," Thompson said. NSF officials are giving about $10 million for two optical network test bed research projects during the next three to four years.
Spurred by "applications looming that will require file sizes on the order of a petabyte to be moved around," scientists are working quickly to master end-user photon networking, Wing said. A photon is a discrete packet of electromagnetic energy.
"You need to break out of current general-purpose TCP networking," he said. "What allows you to do this is light pipes, dedicated channels that can give the user completely dedicated bandwidth" in which delays can be controlled and bounded.
Photon networks differ fundamentally from today's electron, transistor-enabled router networks. "Photons carry no charge at all," Wing said. As a result, "there's no such thing as an optical router." That means transistor chips' memory buffering ability is lost. "One second of buffer delay in the optical world is 200 kilometers of fiber," he said.
Ironically, that means scientists are
resurrecting point-to-point telephone switch technology that routers seemingly eliminated. "In a sense, we're taking a step backwards," Wing said. But "we're doing it now in a framework in which the lowest level of service, the equivalent of [digital service] zero, is 10 gigabits."
But rusting Ma Bell switcher technology won't cut it either. "Switching for photons is tough," Miller said. "Photons are very fast. There are physically based systems that can do switching for photons, but they're slow." As a consequence, "you do the switching before the photons get there. ... You send a signal — could be an optical signal, could be electronic — down the pipeline that goes to the endpoint of the optical network and [the signal] tells [the network,] 'Turn this switch on to this specific path,' " he said.
Users of high-end computers will have access to all-optical networks within years, Wing said. And through the usual conduit of universities, optical networking could be "widespread in commercial use in five to 10 years," if not sooner, Nelson said.
"The Internet always surprises us with sooner," he added.
Test bed funding
Two optical network test beds were funded recently by the National Science Foundation. They are:
Dynamic Resource Allocation via Generalized Multiprotocol Label Switching Optical Networks. Three universities, led by the University of Maryland, share a $6.7 million, four-year NSF grant awarded in September 2003. Along with scientists from the University of Southern California and George Mason University, Maryland researchers "will model many of the functions of an interregional, national or even global wavelength-based research and education network," NSF officials said.
Circuit-Switched High-Speed End-to-End Transport Architecture. Led by the University of Virginia, four research entities received $3.5 million in January for a three-year effort aimed at supporting a broad class of e-science projects. The resulting optical network infrastructure should be capable of "on-demand provisioning of end-to-end high-speed circuits; stable transport to sustain control and streaming operations; and middleware and application software to support data transfers, visualization, steering and control," NSF officials said.
Source: National Science Foundation