MCI WorldCom's Wimmer Lays Out Optical Roadmap
Expansion history
To meet the needs of its share of 200 million Internet users, MCI Worldcom's UUNet division has expanded its backbone 200 times since year-end 1995, Wimmer reports. Traffic and capacity growth is not only driven by the Internet, he notes, but also revenue-rich data services. Internet protocol services in the MCI WorldCom network have grown 56% annually, Frame Relay and asynchronous transfer mode (ATM) 24%, and private line 1%.
The MCI WorldCom network as of 1998 took 15 years to build. Using that as a benchmark and overlaying the build time with the projected rate of network performance growth, Wimmer calculates that the number of days it will take to build the existing network starts at 242 this year, drops to 31 days next year, and shrinks to 0.1 by 2004.
Forecasting major network node demands in five years, Wimmer calculates 1926 router ports, 154 Tb/s router rate, some 1,926 tributary and 5,322 line optical cross-connect ports, and 457 ports of transit traffic—all running at OC-768c (40 Gb/s).
Technology needs
Meeting capacity demand remains the top of optical network business driver, Wimmer says. Reducing central office congestion presents another challenge—minimizing network layers and number of network elements, and conserving floor space and power consumption. Lowering lifecycle cost becomes critical with shortening product life cycles—a six-month lifetime is fine if you don't spend much money on the product, he notes, but not millions of dollars. Service mix flexibility, rapid and large service provisioning, and restorable, configurable, observable, end-to-end network management round out the list.
Bit rate and protocol transparency top Wimmer's list of optical network attributes. Without transparency, he reasons that electronics have trouble keeping up with escalating optical bit rates, and multiple protocols can incite protocol conversion bottlenecks. Despite the promise of next-generation routers and protocol aggregation boxes, "for the next several years, ATM [asynchronous transfer mode] looks like a good aggregation vehicle on the network edge," Wimmer maintains.
Wimmer also seeks extreme scalability—not just in wavelength, but also number of ports, software/routing capacity, and density. Another desirable feature is rapid, flexible, capacity and service provisioning. Wavelengths should no longer be something difficult to deal with—they should be a day-to-day part of the business, he says. To procure equipment with all of these attributes, vendor interoperability remains a stalwart requirement in Wimmer's network. "No one vendor can do it all. Even if it could, we'd choose a second," he says.
Equipment list
The technology challenges that Wimmer poses for vendors include optical transport equipment that delivers 3,000 km reach, 4,000x4,000 port cross-connect functionality, and tunable transmitters. "I think you'll see the first one show up this year," he says of tunable lasers. He also seeks 40 Gb/s and 80 Gb/s time division multiplexing (TDM) systems, with polarization mode dispersion (PMD)-tolerant receivers, optical regeneration, and 100-channel DWDM. He proposes using the C, L, and S optical amplifier bands, as well as distributed amplification to overcome the limits of high-power levels required by high bit rate transmission.
Wimmer seeks routers with 10 Gb/s and 40 Gb/s interfaces and SONET management functions. "If I had my way, I'd put 40-gigabit ports right on terabit routers right now," he says, rather than use 2.5 Gb/s interfaces available now, and 10 Gb/s interfaces just becoming available. To connect these devices, he seeks low-cost, intra-nodal optical interconnects. With tens of thousands of ports inside a central office, it's not feasible to pay for high-performance long-reach optics to connect network elements that are 200 m apart, he explains, calling for interfaces that cost a couple of thousand dollars.
On the transport side, Wimmer notes that in the metro network, "we're deploying 1,000 OC-192 systems per month into our network today." MCI WorldCom has also tried out encapsulated gigabit Ethernet over SONET OC-192, he reports. In long-haul terms, the average transmission section in the MCI WorldCom network now stands at 3,000 km. "We need systems with much longer reach," Wimmer says.
Mesh vs. ring
MCI WorldCom plans to finish deploying IP-over-optical ring architectures this year, using OC-192c interfaces on routers and 1+1 optical ring products and DWDM transponders. When the software becomes available, MCI WorlCom will evolve the network to using optical cross-connects and mesh architecture. Both ring and mesh restoration capability are in the carrier's deployment plans, because neither scenario presents a clearly superior cost/performance proposition, Wimmer says.
Wimmer maintains that mesh restoration may well be the best solution, but that the economic case has not proven out to be superior to ring restoration. By the time you design the capacity needed for mesh restoration, endure the network complexity, and overcome reliability concerns, the simpler ring architecture looks attractive. He's also skeptical about whether the cross-connect algorithms can scale to the necessary port counts, or whether the network will have to be partitioned to accommodate them. "There are a lot of trade-offs there," he points out. "We'll start with what we can do with a high level of confidence and go from there."
"We've been successful at driving the cost out of long-haul transmission," Jackson notes. He says that line cost used to comprise 60% to 70% of transmission cost, with nodal cost representing the remainder. Now line cost only represents 30% of the cost, and nodal cost 70%. "I can afford to burn more line if it saves me cost in the node," he explains. Nodal cost would have to drop to sweeten the mesh proposition, he concludes.
By Erik Kreifeldt