Throughout this book the approaches taken to system design and performance evaluation are based on the constraints of the enabling technology. Available fiber capacity is assumed to be limited by the constraints and imperfections of optical transceivers, amplifiers, and cross-connects. These constraints affect maximum available spectrum, wavelength spacing, and maximum bit rates per channel. Optical connections are assumed to have limited reach, both geographically and in terms of the number of optical cross-connects they may traverse. Sizes of switches as well as their speed, complexity, and functionality are also assumed to be limited by cost and performance constraints, ultimately going back to the limits of the underlying technologies. Trade-offs between optical and electronic methods of implementing connectivity and routing are suggested, in which the optimal design point depends again on relative cost and performance of the enabling technologies.

Although emphasizing that these technological constraints are paramount, we purposely keep as much of a separation as possible between the architectures discussed in the book and the limitations of any specific technology. The reason is obvious: Today's technology is likely to be obsolete tomorrow. After more than a decade of gestation in the laboratory, photonic and optoelectronic technology has matured to the point where a wide range of technological choices are available for implementing each function in a network, so that cost-effectiveness and viability in the field are the primary issues now rather than proof of concept, which was the issue in the network testbeds just a few years ago.

The first edition of this book was published when optical networks were just emerging from the laboratory, mostly in the form of government-sponsored testbeds. Since then there have been fundamental changes in many aspects of optical networking, driven by the move from the laboratory to commercial deployment and by the twists and turns of the world economy. The investment climate in which optical networks have developed has had two major swings as of this writing. During the technology bubble that began at the end of the 20th century, investment in research, product development, and network deployment increased enormously. The activities during this time of euphoria produced advances in the technology base that would not have been possible without the extraordinary momentum of that period. At the same time, commercial network deployment provided a reality check. Some ideas that were pursued in the late 1990s dropped by the wayside because they did not meet the test of commercial viability, and new ones came along to take their place. When the bubble burst after less than a decade of “irrational exuberance,” the pendulum swung the other way. Investors and executives who a short time earlier thought the sky was the limit now wondered if demand would ever materialize for all of the fiber capacity in the ground. At this writing a more reasoned approach has taken hold; that seemingly elusive demand has materialized and, hopefully, a more rational and sustainable growth period will ensue.

Ultimately, the performance of a network is limited by the quantity and functionality of its physical resources. In this chapter we examine the various functions performed in a multiwavelength network, emphasizing the role of the optical resources (located in the physical layer of Figure 1.3) in providing connectivity and throughput. For the most part we use the terms transparent optical, purely optical, and just optical interchangeably to refer to entities in the physical layer. The implication is that there is a clean break between the underlying technology and functionality in the physical layer and that in the logical layer. The physical layer contains optical components executing linear (transparent) operations on optical signals, whereas the logical layers contain electronic components executing nonlinear operations on electrical signals. In reality, as mentioned in Chapter 1, the picture in real networks is more nuanced. For example, some simple signal processing (either electronic or optical) may be present in the physical layers of today's networks, making them “opaque” to a greater or lesser degree. Conversely, as optical technology for signal processing matures, it is beginning to make its way into the logical layers. Nevertheless, the somewhat simplified view of a transparent (linear) optical layer underlying an electronic (nonlinear) logical layer is very helpful in providing a generic model for most multiwavelength networks. It will be used throughout this book, with exceptions duly noted as they appear. To provide a proper framework for the discussion that follows, we start in Section 2.1 with a description of layers and sublayers of the multiwavelength network architecture.

At various points in the book, we use stochastic traffic and queueing models to represent the behavior of a network under conditions of random demand. These are based on Markov processes as well as some more general queueing models, which are summarized in this appendix. A readable and comprehensive treatment of these models may be found in [Kleinrock75].

Random Processes

Random processes, such as connection requests, contents of packet queues, and so forth, can be described as sequences of random variables, often called the states of the process, with state transitions occurring at successive (isolated) time points. (Between state transitions, the state remains constant.) In discrete state processes, the states take on discrete (typically integer) values, whereas in continuous state processes the states take on a continuum of values. For example, a discrete state process might be the length of a packet queue, whereas a continuous state process might be the random noise generated in an electrical circuit. In discrete time processes, the transitions are spaced regularly in time so that a complete description of the process is given by the state sequence alone. In continuous time processes, the transitions may occur randomly, at any point in time.

A realization of a random process is a specific sequence. In the case of discrete time processes, a realization is completely specified as a sequence of states. In continuous time processes, the transition times must also be specified.

In Chapter 5 we discussed shared-channel networks, and the emphasis was on satisfying traffic requirements on a static, multipoint physical topology (a broadcast star or its equivalent). The traffic requirements were expressed in terms of flows on logical connections (LCs), and satisfaction of these requirements involved multiplexing and multiple access to share the available channels efficiently. When combined time and wavelength division techniques were employed, the optical connections supporting the LCs were set up and time shared by rapidly tuning the transceivers over a given set of wavelengths. Because all optical connections shared a common broadcast medium in a static configuration, all optical paths supporting these connections were permanently in place. We now move on to optical connection routing and wavelength/waveband assignment – issues that were absent in the static case. We treat both point-to-point and point-to-multipoint (multicast) logical connections.

Introduction

In this chapter we focus on the optical layer of the architecture shown in Figure 2.1(a); that is, we treat purely optical (transparent) networks with reconfigurable optical paths, in which reconfiguration is achieved by space switching together with wavelength and/or waveband routing. Unless otherwise stated, we assume that there is no wavelength conversion in these networks, so the constraint of wavelength continuity is in force. The earliest proposals for wavelength-routed networks (WRNs) appeared in [Brain+88] and [Hill88].

In much of the subsequent work on these networks, a recurring issue has been to determine the number of wavelengths required to achieve a desired degree of connectivity as a function of network size and functionality of network nodes (e.g., static wavelength routers, static wavelength interchangers, or WSXCs).

Since the beginning of the 21st century there has been a burgeoning demand for communications services. From the ubiquitous mobile phone, providing voice, images, messaging, and more, to the Internet and the World Wide Web, offering bandwidth-hungry applications such as interactive games, music, and video file sharing, the public's appetite for information continues to grow at an ever-increasing pace. Underneath all of this, essentially unseen by the users, is the optical fiber-based global communications infrastructure – the foundation of the information superhighway. That infrastructure contains the multiwavelength optical networks that are the theme of this book.

Our purpose is to present a general framework for understanding, analyzing, and designing these networks. It is applicable to current network architectures as they have evolved since the mid-1990s, but more importantly it is a planning and design tool for the future. Our approach is to use a generic methodology that will retain its relevance as networks, applications, and technology continue to evolve.

Why Optical Networks?

Since the fabrication of the first low-loss optical fiber by Corning Glass in 1970, a vision of a ubiquitous and universal all-optical communication network has intrigued researchers, service providers, and the general public. Beginning in the last decades of the 20th century enormous quantities of optical fiber were deployed throughout the world. Initially, fiber was used in point-to-point transmission links as a direct substitute for copper, with the fibers terminating on electronic equipment.

Optical networks as described in previous chapters of this book have progressed steadily since the mid-1980s from point-to-point transmission systems to broadcast stars to ring networks to fully reconfigurable multiwavelength mesh networks utilizing a wide range of optical layer equipment: reconfigurable add/drop multiplexers, optical cross-connects, optical amplifiers, and optical access subnets. The next frontier in optical networking is the optical packet-switched network.

The present generation of multiwavelength optical networks are circuit-switched in their core, meaning that they are connection oriented. In these networks, regardless of the specific scheme used to set up an optical connection, a significant delay (typically of the order of milliseconds or more) is always incurred during the setup period, during which the intermediate switches between source and destination are configured to support data transport. This means that circuit switching is efficient only when the average duration of the connections is much longer than the setup time; i.e., seconds or more.

In current applications, exemplified by Internet browsing, a typical source will transmit data in short bursts (on microsecond timescales), possibly changing destinations with each burst. This is completely incompatible with the circuit-switched approach, where a source-destination pair holds a dedicated connection for an extended period of time. Of course, the currently accepted solution to this problem is to maintain the circuit-switched optical infrastructure and provide a packet-switched logical layer (e.g., a network of electronic IP routers) over the optical layer to deal with the bursty traffic.

Despite the fact that optical fiber communications has been an active area of research since the early 1970s and optical transmission facilities have been widely deployed since the 1980s, serious activity in optical networking did not reach beyond the laboratory until the 1990s. It was in the early 1990s that a number of ambitious optical network testbed projects were initiated in the United States, Europe, and Japan. Although the testbeds were largely government financed, they planted the seeds for subsequent commercial developments, many of which were spin-offs of the testbed activities. The commercial ventures benefited from the knowledge accumulated from the testbeds as well as from the burgeoning worldwide demand for bandwidth. As a result, multiwavelength optical networks are deployed today in metropolitan area as well as wide area applications with increasing current activity in local access as well. In this chapter we give an overview of current developments in metropolitan and wide area networks. Recent developments in access networks were discussed in detail in Chapter 5. The chapter begins with a brief discussion of the role of business drivers and relative costs in creating the current trends. This is followed by a summary of the early testbed projects in the United States and Europe, which provides the context for a description of current commercial activity in multiwavelength metro and long-haul networks. We continue with a discussion of new applications and services made possible by the unique features of intelligent optical networks, and conclude with some thoughts for the future.

In this chapter we explore the structure, design, and performance of purely optical networks with electronically switched overlays. These are the logically-routed networks (LRNs) that were introduced in Section 3.5. Typical examples of LRNs are networks of SONET digital cross-connects (DCSs), networks of IP/MPLS routers, and ATM networks carried on a SONET DCS layer. To provide maximum flexibility, the LRN should be carried on top of a reconfigurable optical network. Although we generally refer to the underlying infrastructure as “purely optical” (that is, transparent), we shall, from time to time, relax that requirement to include optical networks having some degree of opacity on their transmission links.

Introduction: Why Logically-Routed Networks?

The rationale for using logical switching on top of a purely optical infrastructure has been discussed at various points throughout the book. The number of stations in a purely optical network cannot be increased indefinitely without running into a connectivity bottleneck. The sources of the bottleneck are the resource limitations within the network (fibers and optical spectrum) and within the access stations (optical transceivers).

Figure 7.1(a) illustrates the bottleneck in a purely optical network. Network access station (NAS) A has established logical connections (LCs), shown as dashed lines in the figure, that fan out to stations B, C, and D. If this is a wavelength-routed network (WRN), each LC is carried on a separate point-to-point optical connection; that is, three optical transceivers and three distinct wavelengths are required (assuming that the stations have single fiber pair access links).

The multiwavelength network architecture described in Section 2.1 contains several layers of connections. By exploiting the various alternatives in each layer, it is possible to produce a rich set of transport network configurations. This chapter explores how a desired connectivity pattern can be established using the combined functionality contained in the various layers. The approach is to examine the properties of different classes of networks through a sequence of simple illustrative examples. The design objective in each example is to provide a prescribed connectivity to a set of end systems. Each of the network classes illustrated in this chapter is discussed in more detail in later chapters, as is the issue of optical network control.

Our first example is shown in Figure 3.1. Five geographically dispersed end systems are to be fully interconnected by a transport network, which is to be specified. The end systems might correspond to physical devices such as supercomputers that interact with each other, or they may be gateways (interfaces) to local access subnets (LASs) serving industrial sites, university campuses, or residential neighborhoods.

In all of these cases, a dedicated set of connections is desired (shown as dashed lines in the figure), providing full connectivity among all the sites. Figure 3.2(a) shows one possible transport network, whose physical topology (PT) is a star, in which the central node is a star coupler of the type shown in Figure 2.7(a). Each end system is connected to the star through its own network access station.

In Chapter 2 we proposed a layered view of the connections in an optical network, focusing primarily on issues associated with optical layer transport but including a discussion of transport in logical network (e.g., IP network) overlays as well. Then in Section 3.1 we encountered a different way of “slicing” the functionality of an optical network, distinguishing three planes: transport, control, and management. In general terms, the transport plane is responsible for the physical transfer of data across an optical network, the control plane provides the intelligence required for the provisioning and maintenance (e.g., failure recovery operations) of a connection, and the management plane provides management services such as performance monitoring, fault and configuration management, accounting and security management. This chapter provides a summary of the current state of optical network control, which is a broad and rapidly evolving subject. The reader is referred to texts completely devoted to the subject of control (e.g., [Bernstein+04] for a more comprehensive treatment).

The line between management and control is not clearly defined. But roughly speaking, management functions deal with long-term issues and operate on slow timescales, whereas control functions are associated with rapid changes in network configurations and operate on short timescales. For example, the repair of a network fault such as a cut cable would be a management function. It might require days or weeks. On the other hand, “point-and-click” provisioning, where a network user controls the provisioning and configuration of a connection, is a control function.

Graph and hypergraph terminology has evolved over the years. The following definitions are adapted from [Berge89, Bermond+97, Chartrand+96]. Some of the material in this appendix is found in other parts of the book. It is repeated here for convenience.

Graphs

A graph G consists of a set of vertices V(G) and a set of edges E(G), where each edge e is a pair of distinct vertices (u, v). (If the two vertices are the same, then the edge is a loop. We rule out these cases.) A graph with vertex set V and edge set E is typically denoted by G(V, E). If e = (u, v), then u and v are adjacent vertices and e is incident on u and v. Two edges are adjacent if they are incident on the same vertex. Nonadjacent edges or nonadjacent vertices are called independent. A set of pairwise independent vertices of a graph G, which is of maximal cardinality, is called a maximal independent set. Figure A.1 shows an example of a maximal independent set of vertices (outlined in dashed circles).

A graph in which every two vertices are adjacent is called a complete or fully connected graph. The complete graph with n vertices is denoted by Kn. Figure A.2 shows K5.

A graph G is called bipartite if its vertices can be partitioned into two subsets, V1 and V2, (called partite sets) such that every edge of G joins a vertex in V1 to one in V2.

The Internet has become an integral part of our society. It not only supports financial transactions and access to knowledge, but also cultivates relationships and allows people to participate in communities that were hard to find before its existence. From reading the news, to shopping, to making phone calls and watching videos, the Internet has certainly surpassed the expectations of its creators.

This vital organ in today's society comprises a large number of networks, administered by different authorities and spanning the entire globe. In this book, we will take a deep look into what constitutes those entities and how they form the skeleton of the Internet. Going beyond the physical infrastructure, the wires and boxes that make up the building blocks of today's Internet, we will study the procedures that allow a network to make optimal use of those building blocks.

As in traditional telecommunication networks, the Internet constituents rely on a team of network designers and managers, for their design, operation and maintenance. While such functions are well studied within the traditional telecommunication theory literature, the Internet imposes constraints that necessitate a whole new array of methods and techniques for the efficient operation of a network. Departing from the traditional circuit-switched paradigm, every unit of information flows independently across the network, from the source to the destination, and thus is much harder to account for.

Since the late 1990s, the research community has been devising techniques that allow the operators of this new kind of network to allocate and manage their networks' resources optimally.