As the optical products guys at Silicon Labs, a chip company in Austin, Texas, like to say, “There is light at the end of the fiber.” While it’s good news, we probably won’t see that light until 2005. Some predict that the optical fiber business will begin a rebound in 2004. If that happens, it won’t be a significant recovery. But at this point, even that’s a positive indication given the serious decline over the past few years.
The telecommunications business, the primary source of optical fiber revenue, is slowly downsizing. At best, it’s a replacement/upgrade market. There’s a glut of fiber around the world, and most of it will stay dark for the immediate future. But don’t be fooled. Optical fiber still has life, as it’s the fastest communications medium on the planet. We can look forward to an expanded fiber-interconnected Internet. And more and more, we’re exploring fiber’s ability to transmit data at the highest of speeds in metro networks, local-area networks (LANs), and even in newer storage-area networks (SANs) and passive optical networks (PONs).
The largest segment of the fiber-optic business involves the Sonet/SDH wide-area networks (WANs) that form the backbone of the long-distance telephone system and the Internet. WANs operate primarily at 2.5 Gbits/s (OC-48), but many are upgrading to 10 Gbits/s (OC-192). Furthermore, capacity in these systems is being increased by incorporating dense wavelength-division multiplexing (DWDM), where multiple wavelengths of IR light carry high-speed data over a single fiber.
Fiber is also used in metropolitan area networks (MANs), where it links LANs to the wide-area networks (WANs) and performs other information-carrying duties. One such example is hybrid-fiber networks used by cable TV companies. More and more fiber MANs are expected in the future, many of which will still use Sonet/SDH. However, some new networks will move to 1-Gbit or 10-Gbit Ethernet (1 GE or 10 GE).
Fiber also is encroaching into the LAN space. High-speed LAN backbones in large corporations or government agencies are opting for 1 GE or 10 GE. While 1 GE copper to the desktop is popular, the backbone of the LAN will move to fiber as the need for speed, distance, and capacity increases.
PONs also will grow in usage as telecom carriers begin to implement fiber to the home, curb, or premises (FTTx). PONs are ideal for the last/first mile connections, because their cost and ease of implementation make them competitive with other broadband strategies.
SANs are another bright spot for fiber. With the burgeoning need for massive storage of data, SANs have ramped up significantly over the years. And there’s no reason for any letup, as an increasing amount of data centers and large organizations implement massive data storage systems. These use Fibre Channel (FC) interconnections, which communicate at 1, 2, or 4 Gbits/s. Also entering the fray will be 10-Gbit/s FC systems. The newer Internet small systems computer interface (iSCSI) that uses IP over 1 GE or 10 GE to link servers to storage systems is another popular arena.
Many mergers, bankruptcies, acquisitions, and other consolidations have occurred during the downturn. The result is a leaner and meaner fiber-optical industry poised for the eventual recovery. With faster chips, improved optical components, and some new innovations, the forthcoming optical systems will be better than ever. They will continue as the fastest medium in the universe.
- No real recovery until 2005. Many see a mild recovery beginning this year. The industry will survive thanks to some increased telecom buying and the hope brought by newer applications. Overall, the downturn and subsequent adjustment will give us a smaller but more robust optical fiber industry.
- Dispersion is conquered. Dispersion is the lengthening of light pulses on a fiber cable due to light scattering induced by polarization and chromatic effects. While it has always been a nuisance, dispersion causes less of a problem at lower speeds. But if 10- and 40-Gbit/s optical systems are to succeed, dispersion must be minimized. This area has received lots of attention during the downturn. Today there are both electronic equalization and optical corrective solutions, as well as specialty fiber that greatly reduces the problem.
- 10-Gbit/S optical Ethernet on the verge. The market for 10-Gbit/s optical Ethernet has been around for a few years, with components available, but switch vendors moved slowly because of the downturn. Now, lots of competitive products are finally available. Even with IT spending flat for the past few years, some companies are ready to enhance and build out their LANs and even launch MANs based on the technology. You can never have too much bandwidth.
- Affordable tunable lasers. Dense wavelength-division multiplexing (DWDM) is a technique that puts up to 200 individual wavelengths of IR light on one fiber. It greatly increases the capacity and speed of existing fiber connections, but at a very high price. Because each wavelength requires its own high-power laser, the cost of such systems is expensive. The solution is the now available low-cost tunable laser. Such lasers can be set to any desired wavelength by applying an appropriate dc control voltage. These devices greatly reduce the cost of new systems and lower the cost of repairs and replacements because only a handful of lasers is required, not hundreds as before. Tunable lasers will make DWDM more practical than ever.
- Metro networks to the rescue. While the telecom carriers recover and continue to delay expanding their existing WANs, some carriers are going ahead with metro networks. These intermediate networks interconnect LANs to WANs and LANs to other LANs. This is one of the few growth areas within the fiber-optical-system sphere. Sonet/SDH systems will continue to be competitive, but 1- and 10-Gbit Ethernet will be used in applications that don’t require the quality-of-service of a Sonet/SDH system to lower costs.
- Positive outlook for PONs. Certainly one bright spot in the optical world is the passive optical network (PON). Interest is climbing in deployment of PONs in metro networks as well as in using them to implement the last/first mile connections to homes and businesses. Fiber to the premises (FTTx) makes sense. It provides high-enough speeds to offer not only super-high Internet access, but also speed to deliver video and VoIP. Broadband PONs (BPONs) use ATM-like packet techniques to achieve speeds of up to 622 Mbits/s downstream and 155 Mbits/s upstream with up to passive 32 nodes. Lower speeds are possible with Ethernet PONs at very low cost. This is a growing sector as telecom carriers protect their turf from further moves by the cable and wireless companies.
- Sonet/SDH is adapting. Sonet/SDH vendors are working to make their equipment more efficient while lowering the cost. Given the telecom downturn over the past few years, Sonet gear sales won’t grow until the industry returns to health and equipment prices become more moderate. Look for the adoption of resilient packet ring (RPR) and multiprotocol label switching (MPLS) to further increase Sonet’s usefulness. Competition from 10-Gbit Ethernet in the metro sector is also encouraging vendors to provide more value. With hundreds of thousands of Sonet rings and point-to-point links in place worldwide, Sonet won’t go away. But its growth depends upon fine-tuning the offerings and value.
- 40-Gbit/s systems lay in wait. You can actually buy a 40-Gbit/s (OC-768) Sonet/SDH system today. But these systems aren’t widespread due to high cost and ongoing developments in chips and dispersion compensation components. Right now, the need for that speed is minimal. But as the industry recovers, these systems should find a place as Internet capacity expands. Also look for 40-Gbit/s Ethernet systems to emerge.
- Network processors will improve optical networks. Network processors are well into their next generation of implementation and have become better than ever. They’re slowly making their way into switches, routers, and other high-speed optical equipment. With improved speed and ease of software reconfigurability, they will soon make optical networks more flexible and affordable.
- CMOS will lower optical component costs. As feature sizes shrink and processing techniques improve, CMOS chips just get faster. They’re now the de facto standard for serializer/deserializer and other chips used in optical data systems. As the feature size drops into the 90- and 65-nm range, further improvements in speed will be achieved. By eliminating or at least minimizing the need for gallium arsenide, indium phosphate, or silicon germanium, costs will drop dramatically, making optical gear even more attractive.