technology & profits
Too smart for your own good?
BY STEPHEN HARDY
First you take Internet protocol (IP) packets and encapsulate them into Asynchronous Transfer Mode (ATM) cells. Then you wrap those cells with Synchronous Optical Network/Synchronous Digital Hierarchy (SONET/ SDH) frames via multiplexing and transmission equipment. Finally, you run those frames across the optical layer through dense wavelength-division multiplexing (DWDM) equipment. By the time IP traffic is safely on its way from one end of the network to the other, it's better bundled than the average six-year-old walking to school on a winter's day in South Dakota. Surely there must be a smarter, more efficient way to transmit signals across a fiber-optic network.
Of course there is, equipment vendors say. Four network layers are at least two too many. First, we'll evict SONET/SDH gear from the network core. Soon, we'll be able to do the same with ATM switches. The remaining IP switch/router layer and the optical layer will combine to provide transport, provisioning, restoration, quality-of-service guarantees, and everything else a carrier could want. How will this miracle of telecommunications science occur? Intelligence—we'll move the capabilities of SONET/SDH and ATM into the remaining two layers.
That's a lot of intelligence to cram into a couple of layers of hardware, but both the switch/ router and optical networking communities are off and running in that direction. Unfortunately, the two parties appear to be racing against each other as much as sprinting together toward a common goal. "I think everyone is going to try to do everything," says Chris Nicoll, director, infrastructure analysis, at Current Analysis Inc. (Herndon, VA). Thus, carriers looking to upgrade their networks—or build them from scratch—potentially could face an architectural dilemma in the next three to five years: How do I get two eggheads who want total network control to work in tandem? Put another way, is it possible for equipment to be too smart?
A lesson in semantics
Cisco Systems Inc. (San Jose, CA) and CIENA Corp. (Linthicum, MD) began the siege against SONET/SDH equipment last spring with the introduction of a concept they called "optical internetworking." The introduction of an OC-48c (2.4-Gbit/sec) interface on Cisco's new 12000 Gigabit Switch Router meant that the IP switch/routing layer could operate at the same speeds as the DWDM equipment sitting on the optical layer. This speed match obviated the need for SONET/SDH aggregation equipment in data-network cores, the two companies reasoned. And since the general traffic mix had begun to tip significantly in the direction of data, particularly IP data, carriers across the spectrum should consider an evolutionary path toward this streamlined architecture, the manufacturers urged.Ascend Communications (Alameda, CA) quickly jumped on the bus with its GX 550 core router, which also carried an OC-48c interface. Meanwhile, each of the terabit switch/router startups began to add similar interfaces to their blueprints as a matter of course.
The optical-internetworking message found acolytes among several of the emerging national carriers in the United States. Frontier Corp. (Rochester, NY) inaugurated the construction of separate ATM and IP overlay networks using Ascend and Cisco equipment, respectively, and hooked both hardware types directly into their optical network via DWDM equipment (see Lightwave, October 1998, page 1). Meanwhile, Enron Communications (Portland, OR) took the concept to its most extreme level by announcing it would service nothing but IP traffic over a network comprising little more than Cisco switch/routers, CIENA DWDM gear, and distributed computing resources.
The optical-internetworking movement has continued to build momentum over the past year, with GTS Carrier Services spreading the gospel to the other side of the Atlantic with its announcement that it will test the Cisco/CIENA combination on its pan-European network. In a nod toward the necessity of relocating SONET/SDH protection capabilities in a 2-layer network, Cisco recently announced its Dynamic Packet Transport (DPT) initiative, which uses the MAC-layer Spatial Reuse Protocol to create ring protection for IP packet transport (see Lightwave, May 1998, page 76). The IP switch/router would serve as the seat of future core network intelligence, Cisco signaled.
Just when it looked like the optical-internetworking army would conquer the field without resistance, a group of optical-networking startups—including Lightera Networks Inc. (Cupertino, CA), Monterey Networks Inc. (Richardson, TX), and Sycamore Networks Inc. (Chelmsford, MA)—launched a counterattack. The startups maintained that IP switch/routers won't provide the wavelength switching and rapid restoration that future optical-networks will require. These intelligent functions would best reside in the optical layer, encompassed in what the companies termed "intelligent optical-network" switching fabrics that would either interface with existing DWDM equipment or incorporate DWDM functions into their hardware. These equipment families offered automatic service provisioning and multiple service-restoration options as well as handling the details of signal transport.
While this intelligent-optical-network vision is a year younger than its optical-internetworking counterparts, emerging carriers have shown interest in this philosophy as well. For example, Williams Communications has already issued an order for Sycamore's equipment. Meanwhile, both Lightera and Monterey claim to have no shortage of carriers interested in serving as beta sites when their equipment is ready to leave the lab this year.
A full-fledged competition between the two technologies looked to be a year or two away. But the rules of engagement changed somewhat when CIENA, the DWDM pioneer of the optical-internetworking concept, acquired intelligent optical-networking startup Lightera. Were the two concepts being merged together? Was CIENA hedging its bets? Or had a member of the optical-internetworking fraternity abandoned the concept?
We can work it out
Questioned at the teleconference held in conjunction with the announcement of Lightera's acquisition, CIENA's Steve Chaddick claimed that optical internetworking and intelligent optical networking were complementary technologies and that the addition of Lightera did not signal a lessening of support for optical internetworking.This sentiment will withstand scrutiny, say analysts. "I don't see it as an either/or. I think it's just different people looking at the problem from different directions or maybe even just using different terminology," offers Dana Cooperson, an analyst at RHK Inc. (South San Francisco, CA).
"They're both trying to do similar things," agrees Nicoll of Current Analysis. "I don't necessarily think they're directly competitive; I think they're somewhat complementary. Again, it's just a matter of where you want to put the intelligence. And I think it's going to end up being in both."
The need exists for more intelligence on both ends of the equation, says Nicoll. "DWDM and optics are still heavily a TDM (time-division multiplexing) -based system. And with that, if you put one bit on a channel or you fill the channel up, it's the same to the system—which can be hugely wasteful," he explains. "We're going to have to manage the bandwidth capabilities on the optical network to a much finer degree. This is really where I think Monterey is coming into play. But, in effect, that's what Cisco's trying to do as well: Give us the whole channel, and we'll make sure we fill it up."
Such duplication of functions could lead to trouble in the future, Nicoll says. "You don't want to get into a situation where both layers of the network recognize a problem and both attempt to respond to it," he offers. Carriers shouldn't have to worry about such a conflict in the short term, he says, because current switch/routers can't recognize and react to network emergencies within the 50 msec touted for the intelligent optical-networking equipment. "It's more of a problem in the future as the Ciscos, Nortels, and Lucents...build in the ability to recognize quickly these problems and then respond quickly to them," he concludes.
Delegating authority
With potential problems on the horizon, carriers should follow the evolution of each concept closely and match the capabilities of the product types to the requirements of their networks. RHK's Cooperson, for example, believes the type of network a carrier already has constructed or has in mind for the future—pure data versus multiservice provisioning—could influence which approach holds the most appeal. The former type of network provider fits the optical-internetworking vision well, she says."They're really looking at it from, 'I've got an IP network, and how am I going to evolve my network?' And the other guys are looking at it as, 'I have multiple services, and how am I going to evolve my network and my equipment as the service mix turns more toward data and IP in particular?' she explains. "So I think that there are going to be different evolutionary paths—which means different solutions—depending on who the customer is."
Eventually, both types of hardware will find their way into the same network, and by that point, Nicoll believes that the two camps will have negotiated a division of responsibilities between the optical and switch/router layers. "I would expect that they will figure out some way to work it out," he predicts. "It might be even as simple as turning off, in the switch/router, the optical-recognition capability. I don't think the DWDM equipment is going to be able to enable, down to the application layer, quality of service and those types of application-handling parameters. So I don't think it's going to be that much of an issue. They're definitely going to learn how to play nice together, because it's a benefit to each of their respective groups to do so."
In Nicoll's view, the optical layer will handle the tasks associated with the physical transport of traffic—"The equipment that is providing that physical path is in the best position to respond to physical-path problems," he says—while the switch/ router layer will handle applications and quality-of-service issues.
Smarter than the average network
Regardless of which approach appears more popular at any given time, both camps appear to be headed in the right direction when it comes to the future shape of fiber-optic networks."At least on the optical-networking level, the more intelligence you can build into your product, the more value there will be to the customer base," says Cooperson. "So being able to manage the wavelengths, and therefore switch them and [perform] various other kinds of actions through software, is going to be a big deal. And being able to hook into whatever the network-management systems are that the carriers are currently running is also going to be critical."
"Networks can't be point-to-point anymore," agrees Nicoll. Given the twin drives toward greater intelligence, carriers had best hope that Nicoll is right that the two parties will eventually reach an accommodation concerning equipment responsibilities—or else carriers may wish that their intelligent hardware wasn't quite so smart.