By Robert Pease
Tunable lasers made their splash on the telecommunications scene a few years back, developed primarily for "sparing," in wavelength-division multiplexing (WDM) applications. The ability to tune the lasers to whatever bandwidth was required alleviated a need to stock lasers for each individual bandwidth. Since then, tunable lasers have been developed for use in bandwidth provisioning and, more recently, wavelength-routing applications geared toward an all-optical-network architecture.
Project HORNET (Hybrid Optoelectronic Ring NETwork), a research program being conducted by Stanford University under the direction of Prof. Leonid Kazovsky, is working on new technologies for using tunable lasers for future metropolitan area networks (MANs). The research is sponsored by Sprint Advanced Technology Laboratories in Burlingame, CA. The researchers hope to determine the feasibility and advantages of an optoelectronic packet-switching MAN.
"At the time we started the project in June 1998, everyone in the WDM world was trying to break capacity records," says Ian White, a Ph.D. student at Stanford's Optical Communications Research Laboratory. "We wanted to be different, so we tried to harvest new technologies for the MAN. Tunable lasers were just becoming commercial products at the time, so we decided to incorporate them into the network architecture and determine two things: if it was even feasible and if there were any advantages in doing so."
White explains that conventional MANs are designed around a traffic model in which all traffic comes into the MAN from the Internet backbone and simply needs to be distributed to the end users. In other words, a MAN is merely a distribution network. However, times are changing, and so is the traffic. Technological changes in Internet-protocol (IP) traffic, such as Napster and other applications designed to keep content distributed throughout the network, are leading to a lot of changes in the traffic patterns currently on MANs.
"We predict that there will be a lot of bursty, unpredictable, packet-based communications between access points on the ring network," says White. "Therefore, we see it necessary to give intelligence to the access points. Intelligence needs to be pushed out toward the end user just as content is being pushed out."
That added intelligence could come in the form of tunable lasers. By incorporating a tunable laser into the access points on the MAN, those access points are able to transmit packets directly to each other. In the conventional model, all traffic is hubbed at the center of the star network. HORNET gives the network a logical topology that looks more like a mesh, rather than the star topology typically used in conventional networks.
Benefits of the HORNET architecture are mainly for the network provider, says White. The traffic demands of the future could cause a lot of strain on the switching equipment at the hubs of conventional, static networks. With HORNET, that strain is relieved because the hub doesn't have to route traffic to the destination any more. The access point is now smart enough to perform that function. Also, since the traffic goes directly from the source to the destination, traffic is actually reduced.
Since point-to-point links are eliminated by HORNET's architecture, Syn chronous Optical Network (SONET) cannot be used. Although White concedes that SONET still has a place in point-to-point long-haul, high-capacity backbones for years to come, HORNET will eliminate it in the MAN, except for cases where there is a long-term connection between a source and a destination in the optical layer.
"Because we tune our transmitter on a packet-by-packet basis between packet transmission, we do not maintain a permanent connection between source and destination," explains White. "In fact, we don't maintain any connection at all. Therefore, all of the point-to-point protocols typically contained in the stack between IP and WDM must be eliminated in HORNET."
Research is still underway for providing the inherent capabilities of protection, restoration, and grooming that SONET so aptly provides. The researchers are continuing to investigate how to incorporate quality of service into the HORNET architecture. Currently, the main issue being faced is bit-level synchronization.
"We have a method for doing this," says White, "and it works. But we're trying to improve on it. It's one of the most difficult aspects of the project. The problem is that in HORNET, packets arrive asynchronously in the receiver. The receiver needs to know when to sample the 1s and 0s. SONET takes care of this. However, in our case, we need to do this bit-level synchronization on every incoming packet. Obviously, to maintain low overhead, it needs to happen nearly instantaneously. This is very difficult. Unfortunately, research has not been done in this area. We hope that changes-and we want to be a contributor."
Another important issue is survivability. Again, the Stanford team has a design on how to solve the problem, although it has not yet been demonstrated. As the project continues, more questions can be addressed and other issues will be researched.
"As you can see, this project could be a mother to many more research projects," says White. "We hope that happens and the rest of the research community attempts to tackle the problems we've come across, such as the bit-level synchronization problem and the Layer 2 and Layer 3 issues."
Successful research in the HORNET project could reap many advantages through the use of tunable lasers. A logical topology that resembles a mesh would reduce the amount of switching equipment necessary at the point-of-presence. Nodes would be more connected and traffic across the fiber would be reduced. Routing algorithms would become more flexible and possible paths between sources and destinations could be significantly increased. Additionally, the use of tunable lasers creates a more favorable and efficient environment for multicast and broadcast transmission than those currently available on conventional rings.
What has been achieved so far by the HORNET researchers? The major breakthrough, which White believes has already been partially achieved, is to see an increase in interest in bringing network functions-mainly packet switching-down to the WDM layer.
"People in the industry claim to be doing this, but they're only taking baby steps," says White. "We are hoping to see vigorous efforts to use optical packet switching or optoelectronic packet switching to provide intelligence to the WDM layer."
HORNET has already had several notable successes. For example, tuning the laser between packet transmissions from one ITU wavelength to its adjacent ITU wavelength was accomplished in 4 nsec using a transmitter based on a tunable laser from Altitun Inc. Faster lasers could bring the switching times down to picoseconds.
Project HORNET is innovative and unique, two critical elements we often associate with new optical startup companies. The final element is whether the architecture will work as designed. White and his colleagues at Stanford, with help from AT&T's laboratories, are continuing their research. The team has already reached one conclusion: the optoelectronic packet-switching MAN is definitely feasible and undeniably advantageous.