Optical Networking: Past and Future

Light-based methods have been used for information transmission for thousands of years.  Probably the earliest method is the fires that have been used for signaling since Biblical times. Recall the famous opening of Aeschylus’ play Agamemnon (c. 458 BC):

I wait; to read the meaning in that beacon light,

a blaze of fire to carry out of Troy the rumor

and outcry of its capture….

Smoke signals have been used for communications for a long time, most notably by Native Americans.  Lanterns hung in Boston’s Old North Church (“One if by land, and two if by sea”) figure prominently in Paul Revere’s famous midnight ride in 1775.  Shortly thereafter Lord Nelson (1758-1805) pioneered the use of flashing lights on ships for communication. An optical telegraph was built in France during 1790s by Claude Chappe, which used signalmen in series of towers between Paris and Lille, 230 km apart.  Signals were relayed using movable signal arms.  It took 15 minutes to send a message, but this was much faster than a rider on horseback.

The idea of guiding light through internal reflections was developed in the second half of the 19th century, though for entertainment rather than communications purposes.  In 1880 Alexander Graham Bell patented the photophone, which used light bouncing off of mirrors vibrated by sound.  However the system did not work well and was abandoned in favor of wire-based telephony.  In 1895 there was an early attempt at television by French engineer Henry Saint-Rene using a system of bent glass rods for guiding light images.  About 30 years later, in the 1920s, use of arrays of transparent rods for transmission of images for television and facsimiles respectively was patented by Englishman John Logie Baird and American Clarence W. Hansell.  In 1956: Narinder S. Kapany of Imperial College in London invented glass-coated glass rod, and coined the term “fiber optics”, though his invention was intended for fiberscopes (devices to look inside body cavities or other inaccessible places).

Optical communications as we know it was born with the development of lasers in 1960, and especially infrared lasers on semiconductor chips by Bell Labs in 1962.  In 1961 Elias Snitzer of American Optical published theoretical description of single mode fibers, which are glass fibers with a core so small it can carry light with only one wave-guide mode.  However at the time glass could not be made sufficiently pure for any type of long distance communication.  But in 1970 Corning Glass researchers Robert Maurer, Donald Keck and Peter Schultz invented fiber optic wire or “Optical Waveguide Fibers”, based on Snitzer’s idea, with glass that was pure enough for long-distance transmission (see Figure 1).

Optical Fiber

Figure 1.  Optical Fiber (Source: Nortel)

By 1972 loss was down to 4 db/km, and all the pieces were in place.  In 1973 the US Navy installed a fiber-optic telephone link on a ship, and in 1975 the US Government linked computers in the NORAD headquarters at Cheyenne Mountain using fiber optics.  In 1977 the first optical telephone communication system was installed, just 1.5 miles long, under downtown Chicago.  Each optical fiber carried the equivalent of 672 voice channels (DS-3).  In 1980 the first long-distance fiber optic link was put into service between Boston and Richmond, and the future of optical communications was assured.

Birth of SONET/SDH

Soon it was realized that standardization would be needed to really deploy large amounts of optical communications equipment, so several organizations set to work on this problem, especially the Exchange Carriers Association, a trade group of telephony providers.  These organizations were focused almost entirely on voice communications, so everything in the new system was designed around digitized voice channels, which sampled analog voice calls at 8,000 times per second and quantized them with 256 levels (8 bits), giving 8000 x 8=64kbits/sec.  The result was what we now know as Synchronous Optical Network, or SONET, around 1985.

SONET was designed with qualities important to the telephony providers of that time: ultra-high reliability, rapid recovery (in milliseconds) from problems, and continuous monitoring of signal quality.  A generalization of SONET was developed by what is now the International Telecommunications Union (ITU), called “Synchronous Digital Hierarachy”, or SDH.  SONET is now considered to be a subset of SDH.  It was expected to have a lifetime of 20-30 years.  With the invention of the pure optical amplifier (an amplifier capable of amplifying light waves without conversion to electricity) in the late 1980s the stage was set for converting the world’s telecommunications to optical/digital.  To facilitate design and implementation of telecommunications systems, a set of seven functional layers was defined by the International Standards Organization (ISO), called the “Open Systems Interconnection (OSI) model”, see Figure 2.

OSI Seven-Layer Telecommunications Architecture Model

Figure 2.  OSI Seven-Layer Telecommunications Architecture Model

     The modern optical/digital long-distance telecommunications system is based on the idea of using light pulses to carry information, rather than electronic pulses over a wire, or radio waves through the air.  The advantage of this system has to do with the frequency of the carrier wave: roughly speaking, the maximum information rate is proportional to the carrier frequency.  For radio waves of frequency 1 MHz, this would mean about 1 megabit/second of information.  For radio waves of 1 GHz, the rate is of course much higher, about 1 Gbit/second.  But light waves have a frequency of about 200 TeraHz – about 10,000 higher.  So the optical system can, in theory, send data at rates 10,000 faster.  The original design for SONET/SDH called for use of pulses of a light wave of about 1310 nm, and until relatively recently this modulation method of switching the carrier on and off to signify a one or a zero, called on-off keying (OOK), remained the standard way of transmitting information in SONET/SDH systems (see Figure 3).

On-off-keying

Figure 3.  On-off keying.  In practice, the electrical pulses have a Gaussian rather than a square shape in order to reduce bandwidth consumption.

SONET/SDH Become the Standard

Structurally SONET/SDH is a layer 2 time-division multiplexed system based on the encapsulation of data in frames (framing is essential for data transmission to enable the receiver to determine when a set of data begins and ends; otherwise all it sees is a meaningless stream of random bits).  The basic SONET/SDH frame consists of 90 columns of bytes, with 9 rows, for a total of 810 bytes (See Figure 4).  The first three columns are transport overhead.  The remainder (87 columns) is the Synchronous Payload Envelope (SPE).  Of this, 84 columns are available for carrying the actual payload data (originally conceived as phone calls), which could consist of 672 separate phone conversations (DS-3 rate).  This frame is sent 8000 times a second (the sampling rate for voice calls), giving a total of 8000 frames x 810 bytes/frame x 8 bits/byte = 51,840,000 bits/second.  This is the basic speed of SONET/SDH; all actual transmission speeds are a multiple of this, and are a combination in some form of this basic frame.  Since SONET/SDH was intended as a backbone system, its speeds are very high.  The slowest speed one can purchase is three times the basic rate, or 155,520,000 bit/sec, known as OC-3 in the US, Canada, and Japan, and STM-1 elsewhere in the world.  This is usually just considered to be 155 Mbits/sec.  Other SONET/SDH speeds are multiples of this by a factor of 4: OC-12 (622 Mbit/sec), OC-48 (2.5 Gbits/sec), OC-192 (10 Gbits/sec), and OC-768 (40 Gbits/sec).  In SDH, these are, respectively, STM-4, STM-16, STM-64, and STM-256 (see Figure 5).  The architecture envisioned was for SONET to provide the high-speed bulk transport, with high-speed switching to be done with Asynchronous Transfer Mode (ATM) switches.

Basic SONET Frame

Figure 4.  Basic SONET Frame

 

SignalBit Rate (Mb/s)Capacity
STS-1, OC-1, STM-051.840*28 DS-1s or 1 DS-3
STS-3, OC-3, STM-1155.52084 DS-1s or 3 DS-3s
STS-12, OC-12, STM-4622.080336 DS-1s or 12 DS-3s
STS-48, OC-48, STM-162488.3201344 DS-1s or 48 DS-3s
STS-192, OC-192, STM-649953.2805376 DS-1s or 192 DS-3s
STS-768, OC-768, STM-25639813.1221504 DS-1s or 768 DS-3s

*Not a commercially available speed

Figure 5.  SONET/SDH Speeds and Capacities

At the time SONET/SDH was developed, most people (if they had a computer at all) were on 300 or 1200 baud modems; 64 kbits/sec was extremely fast, and 40 Gbits/second was really astronomical, even for major carriers.  The assumption at the time was that the entire telephone system would be digitized, with end users replacing their old phones with basic rate ISDN phones.  This arrangement would give them voice and also data at 64 or 128 kbits/second, though no one knew what users would do with that kind of data bandwidth.  In any case, it seemed more than adequate for the foreseeable future.  By the 1990s SONET/SDH equipment was finding its way into backbone networks around the world.  SONET/SDH could also be used to carry raw data, of course, and this was done from the earliest days, but voice dominated.  However in the 1990s the Internet grew from the commercialization of DoD’s ARPANET, and being “online” was born.  Soon everyone had a computer and was browsing the web, internet commerce took off, and data overtook voice as the dominant traffic in networks worldwide.  SONET/SDH was there and handled the increasing volume of Internet traffic.

Fiber optic cables are expensive to buy and install in the ground, so carriers began seeking ways to get more utilization out of their plant.  It was realized early on that since a SONET/SDH system just modulated a single wavelength, there was no reason why another SONET/SDH system couldn’t modulate a different wavelength, and then the two modulated carriers (wavelengths) could both be sent over the same fiber—provided that they could be separated at the end.  In principle, lots of such systems could be made to use the same fiber by having each modulate a different wavelength, then combining (multiplexing) them together.  Thus was born what we now call Dense Wave Division Multiplexing (DWDM), which began to be deployed in the late 1990s (see Figure 6; note that in real systems the “colors”—wavelengths—are all in the infrared part of the spectrum).

Schematic view of Dense Wave Division Multiplexing

Figure 6.  Schematic view of Dense Wave Division Multiplexing  (Source: Cisco)

So around the year 2000, the backbone telecommunications systems of the world utilized SONET/SDH to provide bulk transport, ATM to provide switching, and DWDM to achieve high utilization of the fiber plant (see Figure 7).  There was, however, the so-called “last mile” problem, namely that the connections from the end users to the backbone network were quite slow.  What had not been anticipated was the rapidly increasing demand for bandwidth, primarily data bandwidth, on the part of users.  This resulted in the deployment of xDSL circuits by the Telcos, and cable modems by the cable providers, both of which significantly increased user speeds.  ISDN never got a foothold at the end user level because the equipment was too expensive and too slow (too little, too late).  At about the same time, mobile telephone usage exploded, and makers of mobile phones realized that with the extra CPU cycles they had available in their phones, they could offer users extra features and services.  Thus was born the “smartphone”, and its rapid proliferation further increased the demand for bandwidth.

Basic SONET-DWDM Implementation Scheme in Modern Networks

Figure 7.  Basic SONET/DWDM Implementation Scheme in Modern Networks

That bandwidth increasingly involved graphics and video, which were not well-suited for the ATM/SONET framework, especially since most of the data traffic was IP packets, which had to be fragmented to put into ATM cells.  Ultimately the high overhead associated with the IP over ATM over SONET model caused the demise of ATM, and its replacement by a related technology based more on IP itself: Multi-Protocol Label Switching (MPLS).  Methods such as Packet Over SONET (POS) and Generic Framing Protocol (GFP) were devised to allow IP (and other) packets to be more efficiently carried over SONET.

SONET/SDH Limitations

However the increasing use of Ethernet (another layer 2 protocol) made people ask the question of whether SONET itself could be replaced, with the goal of carrying Ethernet packets directly over an optical carrier (wavelength).  To replace SONET, it is necessary to replace its functionality.  That is not so easy for several reasons.  First, SONET is generally implemented with a ring architecture, in which SONET equipment (usually Add/Drop Multiplexers-ADMs) are connected together in a ring fashion (see Figure 8a).  About five rings are required to span the continental United States.  Each ring is usually bi-directional, with two fibers, each capable of carrying the entire traffic load of the ring.  If there is a fiber cut, the protection circuitry of the ADMs is designed to reroute traffic to the other fiber, and thus there is only a brief (~50 msec) interruption (see Figure 8b).  This creates a highly reliable network. In the US most networks use about five rings to go from coast-to-coast.  Second, SONET does important framing and error monitoring functions, which would have to be duplicated by any replacement.  Third, SONET/SDH is an international standard, with equipment from many different manufacturers able to interconnect (called “Mid-span Meet” by Telcos).

 Basic SONET/SDH Ring Architecture and Ring reconfiguration after fiber cut

Figure 8.  (a) Basic SONET/SDH Ring Architecture.  (b) Ring reconfiguration after fiber cut.  (Source: Tektronix)

There are however some problems with SONET/SDH.  First, building SONET rings is very expensive and time-consuming. And if one portion of a ring needs to be upgraded because of traffic, the entire ring must be upgraded.  Often multiple rings must be upgraded.  Second, the nature of Internet traffic makes mesh architectures more suitable, and in such networks higher-layer protocols such as IP have the ability to route around failures (though this is not as fast as hardware switching in SONET).  Third, changes in service to customers—provisioning new service—is inherently slow because of the many steps needed to be done in each ring connecting the customer’s end points.   Fourth, the maximum SONET speed of 40 Gbits/sec, while still very high, is already becoming a limitation.

Beyond SONET/SDH: Improved Capabilities and Higher Speeds

This situation, reached around 2008, led to the development and deployment of several new technologies.  These new technologies seek to address shortcomings of existing networks.  In general, the movement is toward analyzing and operating networks in terms of control planes and interfaces, with the goal of faster provisioning in a dynamic world.  The problem of rapid bandwidth availability is becoming more acute with ever-increasing demand for bandwidth, rapidly changing applications, and new high-capacity networks (~400 Gbyte/sec).  There is naturally much overlap in solutions.  Often proposals are overtaken by events.  The new technologies include the following: Next Generation Network (NGN), Automatically Switched Optical Network (ASON), Generalized Multi-protocol Label Switching (GMPLS), Optical Transport Network (OTN), also known as ITU Q.709, and 40/100G Ethernet.  At a very high level, the relationship among these technologies is shown in Figure 9.

Approximate relationship of new optical networking technologies

Figure 9.  Approximate relationship of new optical networking technologies

  • NGN: NGN, proposed in 2009 by the ITU, is more of an overarching view of how telecommunications will be provided in the future, with a look at all seven layers of the OSI model. It is essentially a managed IP-based (i.e., packet-switched) network that enables a wide variety of services.
  • ASON and GMPLS: ASON is an improved “Intelligent” optical network that can automatically manage signaling and routing through network, with the ultimate goal a type of “bandwidth on demand” (BoD) for users. ASON grew out of rising expectations for network performance of optical networks beyond SONET/SDH, including fast and automatic end-to-end provisioning, fast and efficient re-routing, support of different clients, but optimized for IP, dynamic set up of connections, support of Optical Virtual Private Networks (OVPNs), and support of different levels of quality of service.  It is generally combined with GMPLS as a control plane, to allow implementation of these functionalities. ASON uses GMPLS signaling protocol to set up and monitor edge-to-edge transport connections.
  • OTN: At the hardware level, NGN and ASON usually rely on the OTN, which provides the layer two functionality of SONET, in particular framing, but in an improved version. OTN offers advantages relative to SONET/SDH, including stronger Forward Error Correction (FEC), more levels of Tandem Connection Monitoring (TCM), transparent transport of client signals (data), and switching scalability. The OTN defines a frame format that “wraps” data packets using a format quite similar to that of a SONET frame, with six distinct layers.  Three of these closely mimic the sublayers (path, section, and line) of SONET/SDH.  This includes Optical Channel Payload Unit (OPU), analogous to SONET path layer; Optical Data Unit (ODU), performing functions similar to the SONET line overhead; and the Optical Transport Unit (OTU), similar to the SONET section overhead.  A fourth layer, Optical Channel (OCh), completes the encapsulation.

At the hardware layer, engineers long knew that coherent modulation schemes, whereby phase rather than amplitude is changed, could lead to improved performance and higher capacities.  But the technology to implement these modulation schemes only became available a few years ago.  But now with SONET/SDH maxed out at 40Gbits/sec, and 100Gbit Ethernet a reality, such exotic modulation schemes are being implemented.  These schemes involve modulating a carrier with one of four phase shifts (usually 45o, 135o, 225o, and 315o).  This is called Quadrature Phase Shift Keying (QPSK).  A given phase shift corresponds to two transmitted bits.  So for example 45o would represent 00, 135o would represent 01, and so forth.  This effectively doubles the transmission rate. Schemes with more phase shifts, such as 8 or 16, can also be used, resulting in a tripling or quadrupling of the transmission rate.  For the highest speeds, polarity shifts can also be employed.  Figure 10 shows the method (Dual Polarity Quadrature Phase Shift Keying, DP-QPSK) used in a recent test bed from the Department of Energy and Lawrence Berkeley Laboratories.

DP-QPSK Modulation Scheme Used Achieve 100Gbits/sec transmission rate in 50 GHz Spectrum

Figure 10.  DP-QPSK Modulation Scheme Used Achieve 100Gbits/sec transmission rate in 50 GHz Spectrum

The advantage of such modulation schemes is that they cram more bits into a given amount of bandwidth, at the expense of requiring a slightly higher signal-to-noise ratio, and very sophisticated synchronous demodulation equipment.  Hardware for these modulation schemes is already available from Ciena and Alcatel-Lucent, and some links have been put into operation.  In 2011 Verizon deployed a 555 mile span at 100Gbits/sec in Europe.   Hardware vendors are already talking of 400Gbit/sec modulation.  This however would require changing the ITU grid spacing for DWDM systems, which is based on the OOK modulation method at 40 Gbits/sec.

The Future

At present, the Internet is once again straining to handle the volume of traffic committed to it, and network providers are seeking new solutions.  SONET/SDH are widely deployed around the world, and are valued for their qualities of speed, quality, and fault detection and recovery.  But this technology is approaching the end of its useful life, and over the next decade much of it will be replaced by newer systems based on ASON/GMPLS, the OTN, and higher-speed modulation techniques.

Editor’s Note: Dr. Tom Fowler, a Principal Member of our Telecommunications Faculty, has 25+ years of engineering, R&D, consulting, and teaching/training experience. He teaches courses on Optical Technologies and IP-Based Networks.  He has worked extensively on the development of large-scale US Government telecom networks and is now involved in planning the next generation of those networks.  He is the author of 100+ articles, papers, and reviews.  He has published a book and translated two books.  He has presented courses and papers in the United States, Canada, Mexico, South America, and Europe. He serves as editor of The Telecommunications Review, a widely-respected annual review of trends, issues, and topics in telecommunications.  His PhD in electrical engineering is from The George Washington University; MSEE is from Columbia University; and BSEE is from University of Maryland, where he also completed a BA degree.