Cross-Layer Design in Optical Networks

Optical networking based on Wavelength Division Multiplexing (WDM) has truly established itself as the preferred technology for ultra-long haul communications and networking. Tremendous strides in optical communications technology have been made over the past few decades: wavelength spacings have decreased to 25 GHz; wavelength counts have increased to more than 100; advanced modulation schemes have increased wavelength rates to 100 Gbps; and distributed Raman amplification has increased the optical reach to more than 1,000 km. Meanwhile, optical switching is becoming more and more widespread with the development of Reconfigurable Optical Add-Drop Multiplexers (ROADMs).

In the 1980s optical communications systems smoothly evolved from single-wavelength transmission systems to multiwavelength high capacity ones due to the advent of the optical amplifiers. 30 years later they similarly migrate to optical networking platforms, thanks to the introduction of Reconfigurable Add/Drop Multiplexers (ROADMs) and/or Optical Cross-Connects (OXCs), in order to serve demands for higher capacity systems with long reach and flexibility (Stavdas A, Core and metro networks, Wiley, pp 382–438, 2010; Ramaswami R, Sivarajan K, Optical networks, Morgan Kauffman Publishers, 1998; Agrawal P, Fiber optic communication systems, Wiley, New York, 1997).

However, when fiber was installed and became the main medium of the long-reach telecommunication systems, fiber impairments and component performance limitations for what we define today as high-capacity systems were not evident. When Wavelength Division Multiplexing (WDM) was introduced as an efficient way of resource sharing for the cost-efficient increase of the overall system capacity, physical-layer impairments manifested themselves as cross-channel effects. Meanwhile, as the advances in high-speed electronics and optical components allowed the use of higher channel bit rates, more intra-channel physical impairments appeared even as single-channel effects. Furthermore as wavelength switched optical networks (WSONs) are emerging, ROADMs and OXCs have added their own crosstalk and filtering effects on the transmission links, making decisions on how to efficiently design such optical systems even more complicated. In this chapter we will discuss the various physical degradations that affect emerging optical networks.

Dynamic impairment-aware routing and wavelength assignment (IA-RWA) applies to dynamic core wavelength-division multiplexing (WDM) networks and refers to the process that is responsible for the computation of lightpaths (LPs) for traffic demands arriving during network operation. During this process, the effect of single-channel and multichannel impairments needs to be taken into account so as to compute LPs with strong quality of transmission (QoT) and not to disrupt the traffic that is already established in the network. In addition to ensuring acceptable QoT, the online application of IA-RWA seeks also to minimize the execution time needed per connection. Minimizing the LP setup time is critical during network operation, and the delay experienced due to the LP computation depends largely on the algorithmic implementation and particularly on the way physical-layer impairments (PLIs) are considered. Finally IA-RWA needs to be supported by a properly enhanced control plane that essentially enables the dynamic impairment-aware networking vision.

Volume discount means that the cost of a resource increases at a rate that is lower than the linear increase of the rate, e.g., the cost of a 40-Gbps transceiver can eventually be 2.5 times that of a 10-Gbps transceiver under steady-state conditions, which is less than the rate increase of four times. This scale-up factor is independent of modulation scheme, i.e., with same modulation scheme the scale-up ratio in cost is 2.5 times from 10 to 40-Gbps and 1.5 times from 40 to 100-Gbps transceivers. However, if different modulation schemes are used for different line rates, the scale-up factors will change depending on the different hardware complexities associated with different modulation schemes.

In this chapter we study the planning (offline) problem in WDM networks. In such networks, the signal quality of transmission (QoT) degrades due to physical-layer impairments, making necessary the use of regeneration at some intermediate nodes for certain lengthy lightpaths. Because of physical effects, routing choices made for one lightpath affect and are affected by the choices made for the other lightpaths. This interference among the lightpaths is particularly difficult to formulate in a planning algorithm, since in this version of the problem, we start without any established connections, and the utilization of lightpaths is the variable of the problem. We present a way to formulate inter-lightpath interference as additional constraints on the routing and wavelength assignment (RWA) problem. We initially present algorithms to solve the problem of planning a transparent all-optical WDM network, that is, a WDM network without regeneration capability. Then, we turn our attention to translucent WDM optical networks and present a series of algorithms that choose the regeneration sites and the number of regenerators required on these sites, solving the regenerator placement and the RWA problem for a given set of requested connections.

This chapter deals with the routing and wavelength assignment and regenerator placement (RWARP) in semitransparent optical networks. In particular, it focuses on semitransparent network dimensioning and control, exploiting a cross-layer approach. The chapter describes the physical model adopted and the deterministic RWARP algorithm proposed for the dimensioning phase, computing the necessary resources of the network. Moreover the deterministic algorithm and a new RWA algorithm based on prediction concepts are proposed to analyze the network response in different conditions, in terms of parameter values of the systems and offered traffic. Results of this analysis show that the deterministic approach performs better in case of perfect knowledge of the impairment parameters, while the predictive one outperforms the former in case of imperfect knowledge, which is usually the case in practical settings.

As can be seen in Chap. 3, a large number of QoT-Aware or Impairment-Aware Routing and Wavelength assignment algorithms (IA-RWA) were designed to minimize blocking in dynamic transparent optical networks. The very vast majority of those algorithms were evaluated through extensive simulations. With time, proposed IA-RWA grew in complexity, actually making their accurate evaluation possible only using full-scale simulations. Full-scale simulations, however, tend to be lengthy for the following three reasons: (a) the growing complexity of the proposed IA-RWA techniques; (b) the increasing complexity of the networks that must be modeled (spurred for instance by the increase in the number of wavelengths that can be routed in the network—note that this could be somewhat offset by the deployment of networks with fewer, higher-capacity channels); and (c) the inclusion of more complex QoT models in the simulations—more accurate QoT models are typically more simulation intensive. In addition, establishing a new lightpath may disrupt lightpaths that are already established through the addition of cross-channel effects, such as node crosstalk or non-linear effects. Such disruption is not desirable in a transparent network, and hence in simulations of IA-RWA the QoT of any lightpath that may be disrupted by the arrival of a new demand should be evaluated. Hence, for each new demand, the QoT of many lightpaths may have to be evaluated. If a blocking rate of 10^{− 5} or less is desired, then the simulation of the arrival of (many times more than) 10⁵ lightpaths is required; if a network operator wants to test an IA-RWA in less than 10 min, then a decision must be reached for each demand in (much) less than 60 ms. This can prove difficult to achieve if QoT is to be estimated accurately.

In transparent (and translucent) optical networks, physical layer impairments (PLIs) incurred by nonideal optical transmission media accumulate along an optical path, and the overall effect determines the feasibility of the lightpaths. If the received signal quality is not within the receiver sensitivity threshold, the receiver may not be able to correctly detect the optical signal, and this may result in high bit-error rates. Introducing optical transparency in the physical layer reduces the possibility of client layer interaction with the optical layer at intermediate nodes along the path. The standard GMPLS protocols used for dynamic establishment of lightpaths in transparent/translucent optical networks suffer from lack of PLI information and optical component characteristics. Hence, there is a strong need for development of techniques that provide PLI information to GMPLS protocols and mechanisms that use this information efficiently to evaluate optical feasibility of lightpaths. Without the development of such mechanisms, it would be impossible to automatically initiate a lightpath establishment from client layers, for example, switch or an IP router or a label switch router (LSR). In this chapter, various impairment-aware optical control plane (IA-OCP) approaches are described and compared, namely, signaling-based approach, routing-based approach, hybrid approach, and PCE-based approach. The properties of these control plane architectures are qualitatively studied. Furthermore, the performance study of two IA-OCP approaches is discussed through extensive simulation results.

Modern society relies on communications networks for every aspect of life, from economic transactions to personal connectivity. Yet the communication infrastructure consisting largely of optical networks remains vulnerable to failures that can occur with no warning, resulting in loss of connectivity over potentially large geographical regions. Disruptions, whether intentional or unintentional, can bring down single links or entire network nodes. Mechanisms must be implemented so that networks can either remain immune to or quickly recover from such failures. Techniques that allow networks to survive a failure are referred to as network protection, and typically include significant redundancy, and thus inefficiency. Algorithms for fast recovery from a failure are called restoration techniques. In this chapter we discuss the protection and restoration of optical networks Affected by potential failures and also physical layer impairments.

The energy consumption in telecommunication networks is expected to grow considerably, especially in core networks. In this chapter, optimization of energy consumption is approached from two directions. In a first study, multilayer traffic engineering (MLTE) is used to assign energy-efficient paths and logical topology to IP traffic. The relation with traditional capacity optimization is explained, and the MLTE strategy is applied for daily traffic variations. A second study considers the core network below the IP layer, giving a detailed power consumption model. Optical bypass is evaluated as a technique to achieve considerable power savings over per-hop optical–electronic–optical regeneration.

Many companies today rely on high-speed network infrastructure for real-time applications to conduct businesses. A network link failure will cause enormous data and revenue loss. Thus, routing of traffic with protection becomes a crucial issue in such networks. In this chapter, we first present some key issues in multilayer multiprotocol label switching (MPLS)-over-wavelength-division multiplexing (WDM) networks. Then, we present integrated routing algorithms for the label-switched path (LSP) protection. Next, we explain multilayer protection schemes for differentiated survivability. Finally, we present an LSP partial spatial protection scheme.

The layered architecture of modern communication networks takes advantage of the flexibility of upper layer technology, such as IP, and the high data rates of lower layer technology, such as WDM. In particular, the WDM technology available today can support up to several terabits per second over a single fiber [9], making networks vulnerable to failures, because a failure for even a short period of time can result in a huge loss of data. The main theme of network survivability is to prevent such data loss by provisioning spare resources for recovery. In this chapter, we focus on the impact of layering on network survivability.

The last few years have seen Internet applications evolve towards remote and distributed data processing, storage and visualisation. These applications termed Grid and Cloud applications require scalable, yet efficient coordination of resources across infrastructure providers that own the heterogeneous network and IT resources. In addition, emerging applications demand more in terms of type and quality of service delivery and place more diverse, intensive and yet unpredictable demands for data and resources. Photonic networks play a crucial role in supporting this ever more network-centric application model. The photonic network with an agile control plane is the transport medium that can provide the high bandwidth, agility and speed to offer end-to-end QoS-guaranteed network bandwidth services needed to perform the required data transfer, which is the core of Grid and Cloud applications.

This chapter presents an overview of Grid and Cloud applications, the underlying photonic network and IT infrastructure that supports these applications and architectural frameworks that facilitate the delivery of end-to-end services in this distributed network-centric model. Orchestrating the application requests across the different providers and heterogeneous resources over the optical network infrastructure is key in this environment. This chapter reports on two models for collaborative resource management deployed either at the service plane or the control plane. It also explores key transport solutions that aims to improve the flexibility and agility of the photonic network.

During the last years, a trend has emerged at universities, large enterprises, government institutions, hospitals and public institutions towards acquiring and deploying their own dark-fibre or wavelength networks as opposed to purchasing bandwidth network services from the traditional operators. These institutions usually follow the condominium model to build and deploy their network. The parties get together in a joint effort to purchase the network equipment and deploy the dark fibre. Each institution gets a subset of the deployed fibre and part of the available ports, proportional to their initial investment. However, each institution manages their resources independently of the other organisations that share the physical substrate. Traditional control plane architectures cannot address the requirements of this type of networks, because they assume a single entity has administrative control of all the network elements in a physical domain.

At the same time, a new set of bandwidth-intensive applications are emerging. e-Science applications, Grid applications and high-definition digital media streaming produce such a big amount of data that often justifies dedicating a network to a single application. In order to efficiently manage the resources, these applications must be able to configure the network in the way it better suits their needs.

In response to these requirements, the user-controlled light paths (UCLP) concept, described in Sect. 2, allows a network to be partitioned in several independent management domains and exposes the network resources belonging to each partition as software objects or services under the control of different users. The UCLP results have evolved into Argia (Sect. 3), which is an effort towards creating a commercial product that can be deployed in production optical networks. UCLP and related works are the precursors of infrastructure as a service (IaaS) applied to networks and are changing the way network pieces are being acquired. For making the UCLP concept flexible to any kind of resources, the IaaS Framework (Sect. 4) addresses the need to have a unified framework using enterprise-grade tools and libraries in which new resources can quickly and easily be created. Finally, Sect. 5 introduces the multidomain provisioning systems and the Harmony implementation, an inter-domain broker solution for providing bandwidth-on-demand services over different administrative domains controlled by different local resource managers.

This chapter presents cross-layer network design and control testbeds developed to evaluate the feasibility and effectiveness of networks employing interdisciplinary techniques from the service layer to physical layer as well as to determine issues that need to be addressed before these networks can be deployed in the real world. Firstly, as service-layer aware network design and control testbeds, coordinated computer/network resource allocation demonstrations across several network domains are presented. Using service-layer aware network design and control techniques, users/applications can request network resources that meet their bandwidth and latency requirements through the management plane. Next, as traffic-driven network control techniques that improve the utilization efficiency of network resources, testbed demonstrations of server-layer path provisioning as well as coordinated UNI-link failure recoveries by using the control plane are described. Physical-layer aware network design and control testbeds that overcome the limitations of transparent and translucent optical networks are presented in the subsequent section. One of important findings obtained through the demonstrations is the granularity mismatch issue between an optical path and user/application traffic, which is addressed by introducing coordinated packet and optical path architecture as a short term solution. In the last section, as a middle term solution of this issue, elastic optical path networks, where the right-sized spectral resource is adaptively allocated to an optical path according to the actual client-layer traffic volume and/or network physical conditions, are presented together with testbed demonstrations in terms of service-layer and physical-layer aware network design and control.

Fiber network architecture building blocks can be applied to create “free-space optical wireless network” over the atmosphere with vastly improved and disruptive characteristics for urban and suburban accesses. The prime application will be terrestrial over short distances such as between top of buildings in a densely populated city, but aircraft to ground or satellites and satellites to ground are also possibilities. Communications over rain, fog, and snow are feasible for short but interesting ranges, up to four times the optical depth (or visibility) of the channel. Thus, availability of ranges of around 1 km is high for the 1.5 μm telecom band (to the point of near all weather), and with additional developments in longer infrared wavelengths (e.g., 3.2 μm) that yield longer optical depths, ranges over foul weather can be extended. For both clear and foul weather, atmospheric turbulence impairs transmission and must be properly dealt with. Temperature fluctuations give rise to many random “weak” lenses, and the resulting diffraction effect causes “hot” and “cold” spots of the optical field in the receiving plane. The receiver aperture typically only collects a small area of that field. Hence, it is not uncommon for the link to suffer dropouts of 10–20 dB for 0.1–100 mS every 0.1–1 S. At high data rates, these dropouts represent a large number of bits. Coding and interleaving are possible as a mitigation technique, but at the expense of very long delays, and thus will not work well with typical transport layer protocols such as TCP, leading to serious throughput loss (up to 99 %). Thus, the telecom-based communication network architecture must be modified to deal with this serious impairment to realize a usable system. The channel can be abstractly modeled as an “on–off” channel. When the received power level is high enough for the receiver to demodulate with low error probability, the channel is considered “on.” When fades due to turbulence are deep and the receiver can no longer demodulate with high reliability, the channel is “off.” A two-state continuous parameter Markov Process can be used to characterize the channel with its transition rates given by the strength of the turbulence and the transverse wind velocity blowing the turbules across the channel. These channel fades will not yield satisfactory performance for any serious network usages, and especially when used as part of a short-range open air high-speed (large delay-bandwidth product) edge network interconnected to the wired Internet. In the Physical Layer, transmitter and receiver diversities can be used to reduce the length, duration, and frequency of occurrence of the fades are possible mitigation techniques. Small, low cost optics can be used in transmitter and receiver arrays with the elements spaced more than an intensity and phase coherence length apart (typically 1–20 cm). For strong turbulence, which occurs frequently during day time, additional innovations in higher layers of the network are needed. In the network layer, when the network is multiply connected, multipath routing will increase reliability at the expense of some network throughput. Since this technique couples the resources of the link layer and the network layer, this chapter explores network connection topologies with maximum number of independent paths for the same array sizes of the transmitter and receiver arrays to maximize path diversity performance. Innovations in the transport layer are also needed. Unless great expense is invested in many diversity paths, there will still be occasional residual dropouts. If TCP is used, as in most applications over the Internet, it will interpret these dropouts as congestions at the routers and trigger “window-closing” and “slow-start” resulting in devastating effects on network throughput. A new mechanism must be used in the transport layer to differentiate fading from router congestion and embedded in the ARQ protocol. There will be the necessity for an integrated physical layer to transport layer architecture.