|
|
|
 |
  |
New Age Optical Players
A new wave of optical equipment aims at driving down the economics of metro, long-haul networks. Sam Masud, Senior
Technology Editor
Even as carriers make the transition from 2.5-Gbps channels to 10-Gbps channels in the network core, optical equipment vendors are at the threshold of delivering 40-Gbps wavelengths. Listen carefully, and you'll even hear some noise about 80- and 160-Gbps wavelengths. On the switching front, the debate over OEO (optical-electrical-optical) and OO (optical-optical) switches has shifted from an us-vs.-them attitude to a more reasonable discussion of the value OEO conversions might bring to the network. If you think the future is all about meeting the challenges of providing ever-increasing capacities in transport and switching systems, think again. Indeed, the somberness of the economic climate that permeates the telecom marketplace has done little to cool the industry's enthusiasm for offering new solutions that redefine the capabilities of optical equipment aimed at the metro, regional and long-haul networks. The products that some start-ups promise to bring to market in the next several months offer not just an incremental improvement over what established and new players are shipping, but in many cases nothing less than disruptive network technologies. Take, for example, Cinta Networks, a San Jose, Calif., company that has integrated DWDM transport and switching into one network element. Cinta's WaveJunction system is intended for metro and regional applications and can also provide long-haul transport for distances of up to 2000 km without expensive regeneration. "Our strength is in solving service providers' economic problems today. By integrating DWDM transport and switching, we can initially take out more than 50 percent of their capex and over 70 percent of their opex. And even when they get a fully populated system, we'll still be in the neighborhood of being 50 percent less [when compared to separate optical transport and switching systems]," said John Vaughan, Cinta's president and CEO. Customers could start using WaveJunction as a transport system and later add wavelength switching within the same node. A key system benefit is that it can be scaled port by port. Interestingly, Cinta uses a combination of tunable lasers and optical fiber switching to provide a fully nonblocking switch that allows any channel on any incoming fiber to be switched to any channel on any outgoing fiber (see Figure 1). The use of tunable lasers lets Cinta use one part number for all the different wavelengths instead of using a fixed laser for each wavelength. According to Cinta, the integration of transport and switching within one network element also means one OEO conversion as opposed to two or more that might be necessary when different network elements are used to provide transport and switching. A WaveJunction system supporting 2.5- and 10-Gbps wavelengths and providing 10-Tbps of switching capacity is currently in lab trials with a customer. "It can scale in short order to 4096 ports, and from an architecture standpoint it can go to 8000 ports," Vaughan said.
Intelligent Networks Opthos' IW1000 system relies on emerging control plane specifications such as GMPLS (generalized multiprotocol label switching) and UNI (user-network interface) to close the gap between the electrical and optical domains. UNI, which is expected to be supported in products sooner than GMPLS, is an Optical Internetworking Forum effort based on an overlay model that treats data and optical networks as separate networks with individual administrative domains. In contrast, the IETF's GMPLS spec is based on a peer model, because MPLS-based routers and optical switches are considered peer devices. At Supercomm in Atlanta last month, Cinta and about two dozen other data and optical networking equipment vendors participated in a UNI interoperability demonstration. While the Opthos platform uses in-band GMPLS/UNI signaling between itself and the devices connected to it, it relies on a proprietary messaging protocol and a separate wavelength for handling signaling between nodes in an Opthos network. "We're saying: 'Tell us where you want the wavelength to go and we'll get it there fast, but what you do with the wavelength is your business," said Joe Parker, the company's co-founder and CTO. Like other optical purists, Opthos believes OEO conversions eat up space, power and add considerable network cost. Opthos' zero-OEO solution is based on an optical switch fabric and can provide instantaneous wavelengths to any port on any node on a 400-km ring. By eliminating OEO conversions, the company promises to deliver 32 protected wavelengths in a 10-RU box that sips power in watts rather than kilowatts. At the same time, there are no upgrade costs if a carrier decides to move from 2.5-Gbps channels to 10-Gbps channels. And although Opthos trusts the optical client device to come in on an ITU-grid wavelength, it will verify that it does; if its not an ITU wavelength, the Opthos system will issue a warning or shut down a signal that starts to drift. Additionally, Opthos monitors the signal's SNR (signal-to-noise ratio) at points where it enters or leaves the fabric and also across the fabric to isolate a problem. "Others rely too much on the fact that they have an electrical access point, so they're not taking the steps needed to fully monitor the system," Parker said. Karl Ma, director of product development for DWDM system vendor OptiMight Communications, does not buy into Opthos' proposition of an optical client interface. According to Ma, a decision by a router vendor to use long-reach lasers rather than, for example, the less expensive, easier-to-make, 1310 nm short reach lasers is not a technology issue but a business decision. "The question is: Can the router vendors make money doing this," Ma said. Like a number of others, most notably Corvis, OptiMight's play is to eliminate OEO conversions in the long-haul network, since OEO conversions substantially add to a system's cost. By OptiMight's calculations, its transport platform is more economical for distances of 400 km to 3000 km than either conventional DWDM systems that require regenerators or systems such as Corvis' that require installation of Raman amplifiers. Long-Haul Economics Although not as far along as OptiMight in its product rollout, Innovance Networks is another ultralong-haul player. The Piscataway, N.J., company promises its product will be in beta by Q4 with initial customer shipments scheduled for late Q2 next year. With $95 million in funding, Innovance is revealing little about its product but has a more benign view of OEOs in the network. The calculation behind Innovance's solution is that the fundamental problem for carriers is moving traffic between city pairs, with the typical connection between them being 2500 km to 3000 km. Because this is the case, Innovance argues that considerable OEO costs are tied up at nodes where approximately 70 percent of traffic is merely waiting to transit. Innovance believes it can deliver wavelengths between city pairs and eliminate what it considers hidden OEO conversions. "One issue in building an all-optical network is building the right balance of photonics and electronics. While we will be leveraging photonics technology, we are not going to do so at the expense of delivering against network requirements," said James Frodsham, Innovance's COO. According to Innovance CEO Peter Allen, the company's solution will include transport, switching and an associated photonic control system that eliminates unnecessary, costly OEO conversions, while enabling wavelength provisioning across the network as easily as a DS0 circuit. Although it integrates transport and switching in the same solution, an open architecture permits carriers to use the Innovance transport system with another vendor's switching system. Perhaps less coy than Innovance regarding product details, Maynard, Mass., start-up PhotonEx claims it will solve the problem of delivering fat channels that can also go the distance-from hundreds of kilometers to thousands of kilometers. Touting expertise in optics and RF technologies, PhotonEx's backbone solution will start out as a transport system that supports 40-Gbps wavelengths (the initial system will provide 3.2-Tbps of capacity), but is scalable to 160-Gbps channels. Yet PhotonEx will not just be about delivering big pipes, because the vendor views wavelengths as already commoditized. Instead, PhotonEx talks about using the large bandwidth pool enabled by its technology to create dynamic streams via its IntelliCore system software. This, according to Philip Francisco, the company's vice president of marketing, will allow carriers to deploy bandwidth in "streams," or increments, to meet a specific customer's bandwidth needs, rather than simply providing a wavelength service regardless of whether the customer needs a full 2.5- or 10-Gbps wavelength or not (see Figure 2).
Although PhotonEx, like other optical equipment vendors, will support GMPLS, the company is interested in leveraging the standard to further open its control plane for even more meaningful communication that would allow devices at the periphery of the backbone to provision and turn up services rapidly across the network core. "We've worked with several vendors that build edge devices to extend some features of that standard so more information is known about the relative health and well being of the core network-for example, how busy and even how long the various links are," said Francisco. The system's full capabilities will be rolled out in three phases and include a switching capability that company officials are reluctant to discuss. They said they've spent considerable time with service providers to develop an advanced element management system. "The EMS will give providers the ability to turn up services rapidly and also help reduce the cost of system ownership. I'm talking about things like being able to interrogate the network remotely from the NOC to understand where the circuit cards are, down to the inventory level. So when it's time to do quarterly capital accounting, they know where their capital has been deployed across the network. Today people are sent out with barcode readers to read circuit cards sitting in racks," Francisco said. PhotonEx, which announced at SuperComm that it could transmit 16, 40-Gbps channels across 1500 km without OEO conversions (it used Raman amplifiers in the demo), expects first customer shipments in Q4. Converged Networks
Polaris officials believe the OMX will give CLECs and ILECs the ability to offer the services they desire, while enabling them to replace SONET in the metro core with a high-capacity optical network. Multiprotocol services could be provisioned dynamically on a per-port basis via the OMX, with any port able to handle a mix of traffic with a simple point-and-click operation. Interestingly, in contrast to TDM and optical switch vendors that support STS-1 level grooming, the OMX would support grooming with VT1.5 granularity. "We've visited more than 15 prospective customers and the interest in the OMX is strong," said Sab Gosal, the company's product marketing director. Polaris plans a product launch and trial in Q4. Mahi Networks is taking an equally if not more aggressive approach to the metro/regional network. The Petaluma, Calif.,-based company plans to integrate seven network elements in a converged platform. Mahi CEO Greg Peters would not go into product specifics but said the Mahi system would include legacy transport, optical transport and switching, IP routing and MPLS. Although so-called "god boxes" have yet to take off with service providers, Mahi plans to deliver a carrier-class product with no single point of failure for Tier 1 city applications. "The converged platform is something whose time has come," Peters said. "The legacy equipment providers call it a god box because it replaces four, five, six or seven discrete cash cow products they want to continue to sell. They claim this new equipment is not best in class, but the technologies that have developed from 1999 to 2000 actually enable the replacement of discrete platforms with a chipset," Peters said. Indeed, some carriers have demonstrated considerable faith in Mahi. The $60 million the company raised in the second round of funding included investors such as WorldCom, Williams, Time Warner and Anschutz Investor, the latter a major investor in Qwest. Most unusually, Mahi has no CTO because prospective customers are said to have been proactive in telling the company what they want in the system. "Our customers have been engaged with us for the past 18 months, coaching us on the architecture and the feature release plan. We're oversubscribed for lab trails through late spring [2002]. They know the gold rush mentality of the CLEC buildout is over, and they now have to figure out how they're going to cost out support of legacy services and migrate to next-gen data services at their speed," Peters said. He predicted that Mahi will be in revenue shipments by year-end. No Guarantees At bottom, these new age players are offering tantalizing combinations of price and performance. One caveat is that many have a phased rollout plan for delivering on the capabilities they've outlined. So any slippage would not only bring into question their ability to deliver but, more important, adversely affect their customers. And, as always, the customers will decide which vendors have viable products (assuming the start-ups aren't snapped up by other equipment makers). Given the choices offered by these start-ups, it will take more due diligence on the part of customers before they cast their ballots. Sam Masud is senior technology editor at Telecommunications®. (smasud@telecommagazine.com) |
 |
