Opting for the 'Beauty' or 'Beast' approach to network upgrades
Two different backbone network evolution paths pose critical questions to carriers.
BOB COLLET, Velocita Corp.
The pace of telecommunications networking innovation has truly been astounding when considering where carrier networking was just 10 years ago. Innovation has been going nonstop on many fronts-physical fiber, transmission systems, data networking, convergence, etc., just to name a few. In fact, innovation has been so persistent that network carriers are now facing an interesting proposition: Given limited capital available to upgrade today's networks, where is the best place to start? The answer is far from clear when projected 12 months out.
Looking at backbone networks, there are two different paths possible for upgrading the network. One is increasing the capacity of existing transmission systems (e.g., from today's 10-Gbit/sec to the new 40-Gbit/sec systems that are coming to market), an approach that can be called "bigger is better," or "the Beast." The other path is increasing the network's capacity by increasing the number of wavelengths carried on a fiber. In this approach, optical systems are so highly integrated, they become very small compared with today's system. This approach can be called "smaller is better," or "Beauty."
If a carrier only has so much money to spend, which does it opt for-Beauty or the Beast? So far, both evolutionary approaches have been used somewhat simultaneously in carrier networks. DWDM technology has been deployed on fiber networks in conjunction with the latest and greatest transmission systems. But technology advances are about to take each to extremes that have barely been imagined; optimizing in both areas is either overkill or impractical. Which makes the most sense? Let's take a closer look at the technology implications of each.
Most carriers today, even those fresh with considerable funding, seek to lower the cost of their networks-outright, relative to the amount of customers and traffic carried across the network, or both. There are many factors contributing to lowering network costs (see Figure 1), but the main way that carriers approach that is by squeezing the most transmission out of its existing facilities. Network use, or the ratio of capacity in use over time, is one factor that carriers measure when deciding to add more capacity to the network-either through bigger systems or more wavelengths.
Today's network evolutionary path is primarily that of the Beast. Transmission systems evolve to send more bits per second over a wavelength. Some of the standard carrier network elements and technologies evolving at a rapid pace to support "bigger is better" include 40-Gbit/sec transmission systems, transponders, optical add/drop multiplexers (ADMs), DWDM, optical switches, and software. Several key issues on this path must be addressed, including:
- Increasing wavelengths. DWDM systems increase bandwidth capacity by increasing the number of wavelengths (or lambdas) on a physical fiber. The type of amplification used and the distance between regeneration/amplification locations limit the number of possible DWDM channels.
- Regeneration. Regeneration converts light signals to an electrical state to "clean up" the signal and maintain its original formatting. The cleaned electrical signal is converted back to an optical signal, then transmitted over the next leg of its journey. Typical distances between regeneration stations are approximately every 400 to 600 km. But improvements in laser and amplification technology have extended this distance to 3,000 km. Extending the distance means fewer regeneration stops, resulting in lower costs for running the network. Regeneration is dependent on the grade and condition of the fiber.
- Amplification. New amplification technologies such as Raman amplifiers and erbium-doped fiber amplifiers (EDFAs) take an existing signal and user power to amplify it and transmit the signal over the next leg in the network. Today, amplifiers can support transmission over 60 to 100 km. EDFAs use pump lasers to amplify (or push) the transmission of light signals, while Raman amplifiers reach back 30 km to add a pull effect to the EDFA. The combination helps lower the power requirements as compared to traditional amplification systems and helps reduce the need for signal regeneration. Power is a limiting factor for amplification systems. Amplification requires more power to boost the signal, but using more power introduces unwanted dispersion that, in turn, reduces the number of DWDM channels that can be supported on the fiber.
- Dispersion. Dispersion is the effect of light signals blending together over their transmission distance. Higher levels of dispersion limit the number of DWDM channels.
- Software. Software is the tool that aggregates many customer signals (electrical and optical) onto a network wavelength. This aggregation requires a high degree of electronic processing to inspect packets and ensure appropriate qualities of service for each customer.
Today the choice is fairly simple for carriers. Bigger and better transmission systems, or Beasts, that sit on fiber networks and are optimized using the technologies described here is the only viable choice. Working in conjunction with improved data-networking technologies, the network utilization can be improved in different points of the overall network. But soon, the options for upgrading capacity will change-dramatically.
Increasing wavelengths on a fiber is a viable alternative to adding bigger transmission systems in the network. In fact, depending on the carrier's network today, upgrading at the wavelength level may be more cost-effective than upgrading the transmission systems at the edge of the wavelengths. In the future, it is possible that adding wavelengths could be dramatically less costly than bigger systems.
Figure 2 illustrates the potential of increasing wavelengths on a fiber. DWDM improvements are occurring at nearly two to three times the electronic improvements supported by Moore's Law. The number of wavelengths on a single fiber doubles roughly every six months (at least in the labs). If the current pace continues, by 2005 the number could be in the thousands. By 2010, it could be in the tens or hundreds of thousands. This highly integrated, very dense optical approach we've named Beauty.
For this type of evolution to become reality, the incremental costs of additional wavelengths per fiber must drop dramatically, since carriers won't pay 1,000 times more money for 1,000 times more capacity. This relatively cheap abundance of capacity in the core of the network may significantly change how carrier networks are built. For example, a highly dense, all-optical network would push all the electronic processing to the farthest edge, possibly eliminating those costs or significantly reducing them for the wide-area carrier.
The Beauty approach requires several network changes, with DWDM improvements at the heart, whereby exponential growth of wavelengths can be supported on a fiber. Tunable lasers tune wavelength frequencies to set up an end-to-end path across the fiber network. These devices are critical to using wavelengths efficiently and easily in a very large network.
Dispersion compensators reshape light signals in their optical form rather than require electronic regeneration of the signals. Better signals are produced more quickly than the time required for the conversion and regeneration.
Interleavers are high-end multiplexers that separate and combine wavelengths at any channel spacing and at any bit rate. Interleavers also operate in the photonic domain and serve to make high-count DWDM systems possible. Multiplexing wavelengths together help lower the number of ports required on various system elements.
Liquid-crystal switches support wavelength switching, but without the high port count of micro-electromechanical systems (MEMS)-based photonic switches. Their advantage over MEMS designs relates to the lack of moving parts, which may provide higher reliability while consuming less power.
As these technologies are perfected and brought to market, the building blocks for carriers may shift. By doubling the amount of wavelengths in the network, existing 10-Gbit/sec transmission systems can continue to be used (and since 40 Gbits/sec is arriving, the cost for 10 Gbits/sec will come down).
The overall impact to carrier network design is probably most dependent on when these new "beautiful" systems come to market-and at what cost. If the time frame is "sooner" and the cost "cheaper," then several developments underway today may never gain the critical momentum needed to be market successes. Developments such as:
- Photonic switching. In a beautiful network, tunable lasers, interleavers, and liquid-crystal switches displace the photonic switches that are designed to switch and crossconnect high-bit-rate wavelengths between Beasts.
- Software. Beauty moves intensive electronic processing for multiplexing out to the far edges of the network. Beautiful platforms will likely require less software than Beast-like platforms, which may contribute to reduced development time and cost.
- Switching and routing. Rather than being integrated at Layers 2 or 3, switching and routing functions may occur at the wavelength level. Traditional data-networking switching and routing moves out to the far edge. Again, this is a possible area of cost reduction in beautiful systems.
- Electrical-to-optical conversions. In a beautiful network, these conversions essentially become eliminated from the core and possibly the edge of the backbone network.
- Amplification and regeneration. Dispersion compensation devices could potentially eliminate the need for these two functions altogether, which would save the carrier capital costs for equipment and operational expenses for power.
Although some carriers will use a combination of both approaches, it is more likely each carrier will optimize on the network dimension that supports its corporate strategy. Beauty and the Beast highlight the re-segmentation of the telecom industry that is just beginning to happen today.
An obvious question is how would all the wavelengths be used in a Beauty-based system? First, it should be noted that the wavelength speed would most likely be matched to the end users' end-to-end needs-and it is hard to conceive that speed could ever be beyond 1 Gbit/sec. Therefore, it is not inconceivable that an environment may emerge where wavelengths are as common and ubiquitous as today's voice-grade circuits. Putting these wavelengths to work will require an internetworking capability analogous to how circuit-switched or IP-routed networks are interconnected today.
The good news is that several networking organizations are already hard at work to develop this capability. In addition, it will be necessary for electrical end-user equipment such as IP routers to control the routing of the wavelengths. That's where IP/optical integration becomes useful. Generalized MPLS and optical extensions to interdomain routing IP protocols will enable IP and optical routers to interoperate as peers to control the routing, in contrast to a traditional overlay arrangement whereby each layer is independent.
The peer model will enable IP networks to scale to the dense topologies facilitated by high-wavelength-count networks. In addition, in contrast to the traditional overlay model with each layer requiring its own control plane, the peer model's common control plane would reduce operational complexity and cost.
Carriers have traditionally been segmented based on their geographic coverage (local, regional, long distance) and their customer base (retail, wholesale). For the most part, carriers have sought to provide customers with end-to-end solutions, from the local access and network transport through the data-networking layers. Shifts in the industry are forcing carriers to narrow their focus even further to different "layers" in the overall communications picture.
This phenomenon is highlighted by the new costs to build a network. It is less costly for a new emerging carrier to use the latest technologies to gain the most networking capacity than it is for an existing carrier to adopt and integrate those technologies into its network. Greenfield carriers can offer network capacity at a fraction of the cost for an existing carrier. Legacy carriers are recognizing that and are at a crucial decision point: Do they continue to invest in the physical fiber network and its operations or do they focus on the data/voice networking for their end customers?
The Beast evolution will most likely be adopted by carriers that are providing data services, including those using public IP networks (like virtual private networks) and old-world frame relay and ATM services.
Beauty, on the other hand, will be leveraged by those that own fiber facilities and can support incredibly high wavelength counts. These carriers will optimize their business strategy by providing outsourced networks to the legacy players as well as new customers that emerge to take advantage of the wavelength proliferation.
For the short term, carriers will essentially have their cake and eat it too, using both larger transmission systems and more dense DWDMs to improve their networking capacity. But the technology "fork in the road" will quickly force strategic decisions that will transform the whole telecom carrier landscape.
Bob Collet is chief technical officer at Velocita Corp. (Falls Church, VA). He can be reached at www.velocita.com.