A DISCUSSION WITH ROHIT SHARMA OF ONI SYSTEMS
by Conard Holton
Rohit Sharma is a member of the founding team at ONI Systems (San Jose, CA) and currently serves as chief technology officer, overseeing all aspects of technology development and strategic interests within and outside the company. Before founding ONI, he was with the Telecom Research Labs. He holds Ph.D. and master's degrees in electrical engineering from the University of Alberta, Canada, and a bachelor's degree in electronics and communications from R.E.C. Kurukshetra, India.
WDM Solutions: ONI has established itself as a leading supplier of DWDM metro systems. What are the relative strengths of WDM compared to or used in conjunction with SONET or Gigabit Ethernet?
Sharma: The primary strength of WDM is its inherent flexibility and transparency to carry diverse bit rates and protocols, thereby helping the carriers engineer a true "convergence" layer that can carry any and all forms of traffic and services. SONET/SDH continues to be a primary format for traffic carried over most networks and provides first-generation voice-service transport capabilities. GbE or Fibre Channel protocols are fast gaining popularity as means of transporting data services and we continue to see fast growth in both of these formats in the metro areas. It's imperative that the vendors offer the most open and upgradeable evolution path to service providers so WDM products must be engineered to combine the best of SONET/SDH and optical technologies, and at the same time allow new data services to be introduced without requiring new equipment layers in the network.
WDM Solutions: What optical component technologies are currently key to multiplexing in the metro environment and why?
Sharma: Filters based on thin-film technology remain dominant since they are the most mature technology around for metro applications. While metro core equipment makes use of 100-GHz channel spacings, access equipment can use 200-GHz and even coarse WDM filters. These components are available in volume from multiple suppliers, well-characterized, and easy to design with. In the next 12 to 18 months, additional technologies such as AWGs [arrayed waveguides] will also become relevant in the metro areas.
WDM Solutions: Can you avoid amplification? If amplification is needed, are EDFA amplets strong contenders?
Sharma: Amplification is a critical requirement for providing freedom of traffic-pattern engineering to the service providers. In the absence of amplifiers, carriers can only support service patterns that fit the restricted link budgets and wavelength paths of the system. EDFAs provide the most optimal solution today and since metro applications support multiple channels, wideband, C-band amplifiers with fast transient control are required in the metro networks. Amplets of various types need to provide significantly low cost points to compete with wideband EDFAs and that transition hasn't occurred yet for metro applications.
WDM Solutions: Will SOAs and waveguide amplifiers find widespread application?
Sharma: There is rarely any compromise on performance in optical networks since every fraction of optical power matters. SOAs have long shown promise but the promise has not translated well to practical devices at the right price. I look forward to various SOA manufacturers solving these issues to provide a competitive solution to EDFAs. The ability to amplify the entire C-band at low noise performance and transient suppression are some of the parameters that are crucial to their adoption.
WDM Solutions: Are you offering or planning to incorporate tunable lasers in your systems? Do you believe one type of technology will prevail for tunable lasers or do you even care?
Sharma: We have been evaluating various tunable lasers for almost two years now. Again, a compromise on critical performance parameters such as reliability, output power, wavelength stability, etc., is not acceptable. I think various technologies including multisection DFBs/DBRs [distributed feedback/distributed Bragg reflector lasers], as well as simpler MEMS [microelectromechanical systems] structures, are making progress. On the cost side, they will have to compete with existing fixed or temperature-tunable DFB products.
WDM Solutions: What role will coarse WDM have in the metro marketplace?
Sharma: Coarse WDM is an excellent choice for applications close to the edge of the network where a few channels would suffice and dense WDM for carrying hundreds of signals is typically not needed. There are dual benefits of using coarse WDM in this area—on one hand, the laser transceivers are cheaper because they may not need to be temperature- or wavelength-stabilized, and on the other hand, components such as muxes and demuxes are far cheaper than DWDM components given the relaxed optical requirements, such as 20-nm channel spacing as compared to 0.8 nm for DWDM. I expect coarse WDM to enjoy a fast-growing market in the optical-access marketplace.
WDM Solutions: Do you see role for 40 Gbit/s in metro? When?
Sharma: I think we will first see 40 Gbit/s in the long-haul segment of the network where capacity relief may be more cost-effective at 40 Gbit/s per channel and 80 to 128 channels. On the metro side, 40 Gbit/s will be required when the end devices such as IP routers have 40-Gbit/s interfaces. Given that the distances covered in metro are typically smaller by an order of magnitude, engineering 40-Gbit/s systems for metro will not involve as many impairments as long-haul systems. If various transoceanic cable operators adopt 40 Gbit/s, that will drive the need for 40 Gbit/s in the metro as well. These systems will be for distributing bandwidth to various central offices in metro or regional area from cable landing stations. We may see the first few 40-Gbit/s interfaces on IP equipment emerge in late 2002.
WDM Solutions: Are you integrating all-optical switching into your systems? What technologies look promising?
Sharma: We have been shipping all-optical switching in our ONLINE9000 products since late 1999. We currently use optomechanical switches—essentially big MEMS—and have been awaiting MEMS technologies to mature to integrate more all-optical switching in our platforms. Other technologies such as thermo-optic or liquid crystals have suffered from various issues, including lack of isolation. I think for 8 x 8 to 128 x 128 port counts, 2-D and 3-D MEMS switches of various types will offer some practical solutions soon. Some large 1024 and higher port-count systems will be in lab tests from other vendors this year, and we look forward to MEMS components maturing to offer a viable alternative to optomechanical switches in the next 12 months.
WDM Solutions: Is the market for metro WDM growing fast in Europe and Asia? Any differences in the technologies required?
Sharma: Metro WDM markets are growing fast everywhere since there is a need to offer new revenue-generating services to customers of networks that are in the metro areas. Of course, Europe and Asia have specific characteristics—for example, in Europe, urban areas are quite dense and cities are closer together, so most EU countries are like super-metro regions and require metro-like solutions in the city-to-city network as well. The transparency to formats means that metro WDM systems can support SDH as well as SONET protocols or other region-specific variations in traffic and services, and therefore metro WDM is expected to be a common reality around the world.
WDM Solutions: Finally, how do you work with your component and subsystem suppliers? How much assembly and testing do you do?
Sharma: We have long-term supply and development relationships with most of our component suppliers, and we constantly work with them to help them integrate various functions in subsystems and modules, rather than supplying discrete components to us. At ONI, we are focused on system-level test and verification, and our vendors provide module-level assembly and testing according to our needs and specifications. On the manufacturing side, we have put in place a unique system that allows us to ramp up volume at our optical contract-manufacturing partners by developing our own fully automated test stations, which are then placed at various partner sites and networked in to a common manufacturing system. This allows us to exercise complete control over the test suites and have flexibility in assembling the required systems.