In the boom of the network market, it seemed clear that 40-Gbit/sec technology (40G) was the next step in coping with increased demand. Carriers were rapidly expanding and interested in the increased capacity and subsequent cost savings that the technology could offer.
The current climate, however, has done a lot to slow the deployment of the technology. That is typical of the problems faced today by new technologies looking to gain a foothold in the long-haul (LH) carrier industry. The telecommunications sector has taken a turn for the worse, and speed breakthroughs have been seen as less important than cutting costs and making the best use of currently deployed infrastructure. Many experts have been quick to point that out and write off the 40-Gbit/sec OC-768 (SONET) and STM-256 (SDH) technologies, saying there is enough capacity in the ground already and not enough applications that require faster speeds to drive the technology forward. Implicitly, 40 Gbits/sec will remain dormant until the next major deployments of LH networks.
Although this conclusion is understandable, developments in the future will differ. Many experts believe that LH will be the area to deploy 40G, as it has been with every other major speed improvement. In fact, very-short-reach (VSR) interconnects will likely be the catalyst for industry adoption of 40 Gbits/sec.
For many, the thought of VSR as the first application of 40G might be quite a shock, especially since some network equipment manufacturers have already announced LH equipment. But examining the technology and the economic drivers for 40 Gbits/sec shows us why that may be true.
Cost, not capacity, will be the driver for 40G. Specifically, cost per managed bit, when compared with the alternatives, will determine where and when 40G is deployed. Essentially, the cost of deploying a 40-Gbit/sec channel will be compared to four 10-Gbit/sec channels. Historically, each factor of four increase in speed has been accomplished by a cost increase of less than four (typically 2.5), resulting in 40% cost savings per bit. That was true in the transition from 2.5 to 10 Gbits/sec and is expected in the transition from 10 to 40 Gbits/sec. The component cost directly affects the equipment cost.
Another consideration is that besides the equipment cost, there is the "manage" in "cost per managed bit." By reducing the number of ports by four for the same capacity, equipment can be made smaller, will require less power and cooling, and will reduce the number of fibers by four. All these factors decrease operating expenses. For DWDM equipment, the number of wavelengths to be managed shrinks by four as well, lowering inventory levels for the manufacturing processes and spares. Besides the raw component costs, these advantages are some reasons that 10 Gbits/sec replaced 2.5 Gbits/sec in LH equipment.
Given this scenario, when will 40-Gbit/sec equipment reach a lower cost-per-bit point than 10-Gbit/sec equipment? Some LH equipment vendors are claiming the time is now, and will quote the price points to prove it. Assuming that is the case, why aren't 40G deployments occurring now? In fact, field trials are underway. But another effect now delays 40G.
In the current economic climate, carriers are being very frugal about new deployments. Most carriers already have LH systems with sparsely populated wavelengths. Instead of deploying new systems when they have a bottleneck, they simply add new wavelengths. These wavelengths are at the rate supported by the deployed line systems, usually 10 Gbits/sec. Even if a 40-Gbit/sec wavelength would be priced lower than four 10-Gbit/sec wavelengths, 40G is not an option.
Current LH systems are based on erbium-doped fiber-amplifier (EDFA) optical-amplifier technology and don't have the improved optical signal-to-noise ratio (OSNR) required by 40 Gbit/sec systems. The extra 6 dB of OSNR required by 40G is usually implemented via Raman amplification, which is a fundamentally different system design than existing 10-Gbit/sec systems.
The result is that while it's easy to add capacity by adding more 10-Gbit/sec wavelengths to existing systems, 40G can't be added even though it may be fundamentally cheaper. Given a choice between the addition of more 10-Gbit/sec capacity or the implementation of an entirely new system, carriers are choosing the former in today's economic climate.
Armed with this information, it would appear 40G will indefinitely be delayed until there is another big build-out of LH systems. This assumption, however, would be wrong. The technical advances that have enabled 40G to dip below the cost-per-bit levels of 10 Gbits/sec in the LH sector are now being targeted at VSR interfaces, which are limited to 2-km spans. VSR interfaces are used as interconnects between switches, routers, and transport equipment within a central office (CO).
Technology at 40 Gbits/sec has an even more compelling cost-per-bit position in this application, since the short distance eliminates the need for dispersion compensators, optical amplifiers, and exotic modulators. More important, a completely new system deployment is not required to use 40G interconnects, since each signal uses its own fiber and existing fiber is easily capable of supporting 40G over these short distances.
A key development enabling short-reach (SR) 40-Gbit/sec interfaces is the Serializer/Deserializer Framer Interface Level 5 (SFI-5) standard (see accompanying sidebar, "Standard to drive down costs"), which allowed the development of standardized transponders and the components that go into them for these SR 40-Gbit/sec applications. Complete 40-Gbit/sec transponders are now being sampled, so expect equipment using these components to be available within a year.
Of course, at some point the 40G signals generated at the core of the network through aggregation will have to be transported over a LH system. Logically, it can be assumed that the LH network's limitations would once again delay the transport of these signals. There is an alternative that addresses this problem. A technique known as "inverse muxing" can be used to take a single 40-Gbit/sec wavelength and transmit it over an LH network using four 10-Gbit/sec wavelengths. At the other end, the four signals could be multiplexed back together to give a single 40-Gbit/sec wavelength. In this scenario, 40G exists within the CO as a high-speed interconnect between office equipment, while the existing 10-Gbit/sec systems support traffic between offices.Eventually, with 40G VSR applications acting as catalysts within the CO, it will become compelling to deploy native 40G in the LH network as well. With the current portfolio of 40-Gbit/sec LH equipment expected to grow by the end of 2003, expect 40G to be in the basic capability set of most new transport equipment by then.
Metro networks would seem the next logical step, but evolution there is slightly more uncertain. Metro systems have high cost sensitivity, and bandwidth is usually deployed in smaller increments. But in the metro core as an alternative to WDM or as intermediate-reach links between COs, 40G may have an economically viable role as well.
It's impossible to ignore testing at this point. One of the more important stages of any new technology rollout is the testing phase, and 40G is no exception. However, the problems faced when testing telecom equipment are multiplied when faster speeds are involved. As such, 40-Gbit/sec systems have presented the biggest challenges thus far. These challenges are not only for the
final telecom network equipment, but also the design and testing of the components and modules throughout the food chain. Technology at 40 Gbits/sec is especially challenging since it is advanced in three areas: digital design, microwave engineering, and optics. Besides the testing done within component and equipment manufacturers, test equipment must also be available for the installation and commissioning process before 40-Gbit/sec systems can be deployed.
In this respect, there is positive news (see Photo). For some time, the industry has provided testers capable of making the critical optical and electrical tests such as dispersion, eye-diagram, and bit-error rate. Recent announcements of OC-768 SONET testers now offer the final piece needed to not only test 40-Gbit/sec-capable equipment, but also to deploy 40G in the field.
The advent of 40G in VSR applications is in retrospect not so shocking. Systems at 40 Gbits/sec would be in their first LH deployments today, with SR interfaces emerging, had it not been for the downturn of the telecom industry. Unconstrained by legacy systems, SR interconnects now offer an ideal application for 40G as Moore's law's relentless march delivers superior cost per bit.
Expect to see a number of products announced in 2003 with SR 40-Gbit/sec interfaces. These products, along with new test products that allow 40G to be efficiently deployed, will become the catalyst for the eventual mainstream adoption of 40G.
Larry DesJardin is high-bandwidth program manager at Agilent Technologies (Santa Rosa, CA).
Serializer/Deserializer (SerDes) Framer Interface Level 5 (SFI-5) is a standard electrical interface for 40-Gbit/sec transponders and framer devices that promises to drive down costs even further. The SFI-5 Implementation Agreement was approved by the Optical Internetworking Forum in June.
As a result of the agreement, manufacturers of high-speed SerDes devices and optical modules are able to develop components with the certainty that complementary products from forward error correction and framer suppliers will be interoperable. SFI-5, along with VSR-5 (which defines very-short-reach optical interfaces for 40 Gbits/sec), provides the basis for industry-standard 40-Gbit/sec transponder modules. Economies of scale produced by these standards are expected to drive down prices for 40-Gbit/sec components and systems.