Photonic switching to evolve from WDM crossconnect to ATM switching
Breakthroughs in photonic switching are expected to result in routing and switching applications that combine the economies of scale of multichannel lightwave transport with add/drop capability to meet the network crossconnect requirements in broadband access and Asynchronous Transfer Mode
rod alferness
LUCENT TECHNOLOGIES, Bell Laboratories
The promise of photonic switching is gradually delivering benefits in telecommunications applications. With a basis in space arrays and wavelength-division multiplexing (WDM) technology, photonic switching is moving first to a WDM crossconnect device and then to a rapidly provisioned WDM crossconnect system. From a research perspective, photonic switching technology is moving to a WDM space-circuit switch and Asynchronous Transfer Mode (ATM) cell distribution. All these efforts are expected to ultimately yield an ATM switch augmented by photonics and based on novel architectures and foreseeable technologies.
Recent achievements in using WDM technology have markedly increased the capacity of high-capacity fiber-optic lightwave systems. For example, research teams in Japan and the United States reached new high-bit rates in lightwave transmission technology--a trillion bits per second--earlier this year, well before the predicted turn-of-the-century mark. Teams from Fujitsu, Lucent Technologies/AT&T, and Nippon Telegraph & Telephone (NTT) reported successful terabit transmission experiments in post-deadline papers presented at the 1996 Optical Fiber Conference:
Fujitsu`s researchers transmitted 20 Gbits/sec of non-return-to-zero signals over each of 55 wavelengths through 150 km of conventional singlemode optical fiber.
A research team from Lucent Technologies` Bell Laboratories and AT&T Laboratories used the polarization multiplexing of 25 wavelengths to generate 50 frequencies, each of which transmitted 20 Gbits/sec per second through 55 km of non-zero dispersion-shifted fiber.
An NTT team transmitted 100 Gbits/ sec over each of 10 channels through 40 km of dispersion-shifted fiber.
These impressive results offer further evidence of the growing role of WDM technology in lightwave transmission systems. They suggest that breakthroughs in photonic switching expected in the late 1990s might be the emergence of cost-effective multichannel routing and switching applications. One possibility is to combine the economies of scale of multichannel lightwave transport with add/drop capability to meet the requirement of network-level applications served by crossconnect devices. In the long term, these achievements should extend to applications in broadband access and ATM switching.
To date, commercial lightwave transport equipment using electrical time-division multiplexing (TDM) has evolved from 45 Mbits/sec in 1980 to 2.4 Gbits/sec in 1991 to 10 Gbits/sec in 1996. Total throughput of commercially deployed systems is currently at 20 Gbits/sec per channel using eight wavelengths at 2.5 Gbits/sec per wavelength.
Based on this progress, the cost of transport has fallen sharply. From 1975 to 1995, a 10,000-fold improvement in cost per bit has evolved. At these cost levels, network providers have deployed lightwave systems redundantly and in architectures that virtually eliminate network failure. A key technology driver to this era of strong transport progress has been photonics.
Progress in switching technology, however, has not moved as rapidly. Switching occurs in various models at network facilities. These models include hubs where large volumes of concentrated traffic are routed over long distances, central offices that are linked to individual terminals, and other local area network (LAN) or broadband-access facilities. During the two decades when transport cost fell by four orders of magnitude, the cost of switching has drifted lower by perhaps three to four times.
Switching is an area of opportunity in telecommunications where photonics is poised to make substantial contributions. With its immunity to interference, photonics is ready for application as a broadband switch element and, most immediately, in broadband crossconnect switches in network hubs directing high concentrations of traffic.
Transparency is key
Bit-rate transparency is an important issue to high-traffic applications. One characteristic of photonic switching that has held promise since 1980 has been its transparency, which means that all signals delivered into an input channel pass unaffected into an output channel through the switching point. Transparency has emphatically distinguished photonic switching from electrical switching. In the electrical domain, switching is digital, logical and regenerative. Electrical switches operate at fixed data rates, but very high rates present problems because of power dissipation requirements and interconnection limitations.
As a result of the virtually unlimited bandwidth potential of optical fiber, the merits of a transparent photonic switch coincide with the goal of achieving very high rates of transport. Photonic routing occurs with few limitations on frequency or data rates. Furthermore, WDM technology allows individual routing of different wavelengths, or signal channels.
Multiplexing trio
Presently, multiplexing technologies point the way to wider use of photonic switching in network transport applications. Three multiplexing techniques can be employed to merge multiple lower-bit-rate channels onto a higher-capacity channel: electronic TDM, WDM and optical TDM.
Most deployed transmission systems routinely employ electronic multiplexing to concentrate traffic into channels of traffic up to 2.5 Gbits/sec and, just recently, 10 Gbits/sec. Developers are also demonstrating the layering of channels by means of WDM technology. In addition to providing higher capacity, WDM wavelengths or channels allow the adding and dropping (routing and switching) of wavelengths at connection nodes within a network using relatively straightforward optical elements.
Several market forces, however, are driving toward higher-capability photonic switching. Transport volumes are growing enormously. Point-to-point connections are swelling into the megabit range as voice and data traffic grow with new, high-value applications such as facsimile and online services. Industry analysts foresee the delivery of switched video on a large scale, which could lead to order-of-magnitude increases in network requirements.
Moreover, fiber optics technology is also pushing toward higher-capability photonic switching. The optical-fiber amplifier routinely enables the network regenerator spacing of 80 km rather than the 40 km of electrically regenerated systems. Spans of 600 km between electrical regeneration sites are practical. WDM technology allows order-of-magnitude increases in transmission capacity without the limitations of dispersion in embedded fiber for corresponding high-bit-rate, single-channel systems, such as 10-Gbit/sec OC-192. Network researchers are exploiting the economies offered by fiber amplifier and WDM technologies. They can now switch channels without stepping the signals down to electronic format, often eliminating a layer of network equipment.
Broadband crossconnect
The WDM crossconnect switch differs from the conventional TDM crossconnect switch because it employs a photonic demultiplexer at the input and a photonic multiplexer at the output (see Fig. 1). The demultiplexer separates the incoming wavelengths and outputs each wavelength onto a single channel. Each channel then enters a photonic space switch, consisting of an optical waveguide. In this waveguide, each wavelength is directed appropriately by electronic control to the desired output path.
These WDM crossconnect switch principles are realized in network elements being built by Lucent Technologies to address specifications proposed by the Multiwavelength Optical Networking (monet) consortium. The consortium consists of Lucent Technologies, Bell Communications Research, AT&T, Bell Atlantic, BellSouth, Pacific Telesis and SBC. The consortium is funded in part by the Defense Advanced Research Projects Agency (darpa).
The monet all-optical network encompasses a variety of network applications, such as voice, LAN interconnect and video. The monet crossconnect system is expected to direct WDM-channeled traffic in a transport fabric supporting ATM, Synchronous Optical Network (Sonet) and other equipment types. Within the network layer, monet is going to employ wavelength crossconnection and WDM routing, combined with wavelength add/drop multiplexing. In this ap proach, monet is anticipated to route channels photonically through the network (see Fig. 2).
For broadband services, monet is scheduled to explore the practical capability of switching and rapidly provisioning channels for those services within the transport layer. Advanced new signaling and control systems are required, as well as robust cross connect switch design. This architecture is also capable of sharing resources with the public network infrastructure and providing expected efficiencies and econ omies.
However, issues still remain for wavelength-level provisioning and high-level, rapid service restoration. For example, channel size requires close design analysis. A 2.5-Gbit/sec channel becomes a large channel in today`s network, although it should be come a relatively smaller portion of traffic on a fiber carrying 8 or 16 channels.
Also, as channels increase in number, wavelength conversion to ensure that such a network be nonblocking proves challenging. The problem arises because signals on output channels might derive from signals of different frequencies at the input. Yet, each output channel must consist of only a single frequency. The solution is to convert the signal wavelengths after demultiplexing on the input side of the switching fabric.
A technology to do this would enable extended optical networks and wavelength reuse. It would achieve a strictly nonblocking wavelength crossconnect switch (functionally equivalent to a time-space-time switch) and facilitate development of a wavelength-division switch. Proposals for wavelength conversion seem promising for systems operating from 5 to 10 Gbits/sec. In near-term applications, this challenge may not need a solution. From a network perspective, it is possible to avoid wavelength conversion through a wavelength-selective crossconnect switch based on network management approaches.
ATM switching
ATM technology is also driving photonic development and can handle a range of services. Because ATM is processing- and memory-intensive, however, it challenges photonic implementation. Nevertheless, there is a near-term prospect for photonic processing. To provide terabit-per-second ATM switches, however, ATM switching requirements that appear to conflict with the strengths of photonics must be addressed.
Fortunately, an architecture has been proposed that extends the development of crossconnects and promises to yield a terabit photonic switching fabric in the ATM model. An expandable ATM concept, proposed in 1992 by Kai Eng, Mark Karol and Yu Shuan (Albert) Yeh at Bell Laboratories, features a horizontal cascade capable of receiving 256 optical input lines at 2.5 Gbits/ sec. These lines, or channels, enter a large, highly decomposed fabric within a cell distribution unit. This unit manages expansion and is memoryless and nonblocking; it needs to be reconfigured not at the bit rate but at the slower packet rate (see Fig. 3). It couples a small processing- and buffer-intensive module with a large reconfigurable distribution fabric. The former device accomplishes address decoding and buffering; the latter needs no memory or logic.
On the output side, cells that find a path through the distribution unit enter ATM switch modules that act as packet output modules. The delay and throughput performance are considered well-suited to ATM applications.
As cycle improvements to the tens-of-nanoseconds level are achieved, a photonic ATM switch is expected to become practical. It also has the potential to be an expandable architecture by applying partitioning to the large network fabric, building it from many smaller arrays--even as small as 16 ¥ 16. Consequently, the potential exists to apply space switches that could be applied in modules being developed for WDM crossconnects. u
Rod Alferness is head of the photonic networks research department at Bell Laboratories, Lucent Technologies in Murray Hill, NJ.