This paper describes a new approach for implementing optical switching in a MPLS (Multiprotocol Label Switching) domain where optical packet switching is deployed by using deflection routing. By implementing wavelength multiplexing along with deflection routing, the need for optical buffers is reduced considerably, thereby alleviating of the problems associated with optical packet switching. Using MPLS traffic engineering extensions that naturally facilitate the use of alternative routes furthermore eases the control of deflection routing. The following assumes knowledge of MPLS terminology and functionality.

We base our study on a scalable optical asynchronous transfer mode (ATM) switch using a wavelength-division multiplexed (WDM) and space-division multiplexed (SDM) interconnection network. This switch uses an algorithm to achieve near 100% throughput but the delay is worse than that of an input-queued switch for low loads, and is significantly inferior to that of an output-queued switch for high loads. Another drawback is the difficult hardware implementation. The performance of this modified switch architecture is studied by numerical simulations. Under Uniform and Poisson input traffic, it achieves near 100% throughput and an average delay similar to that of an output-queued switch.

The increasing volume of Internet packet traffic in the telecommunications network has instigated a revolution in traffic modelling techniques, where self-similar instead of Markovian traffic models are now used to produce a realistic representation of data traffic. This paper presents a brief introduction to self-similar traffic and its impact on the performance of an optical node, in terms of packet loss probability and delay. It is demonstrated from simulation results that by employing multiple wavelength channels within the switch and incorporating void-filling, a packet loss probability of 10^-6 can be attained. The generated traffic was subject to statistical tests, and its Hurst parameter was determined using parameter estimation to verify the theoretical assumptions inherent in using infinite variance ON/OFF sources to capture self-similarity.

In this paper we present a control channel-based modular switching technique for packet transport over next-generation high-speed dense- wavelength division multiplexed (DWDM) networks. The proposed technique reduce electronic processing bottleneck in DWDM networks by sending routing information on a dedicated control channel in the form of pilot control packets. The pilot control packets are processed electronically by a control processor at each intermediate node. Processing of pilot control packets ensures that the data packets propagate from source to destination in photonic form (that is, without any processing by intermediate nodes). In the proposed switch, different modules perform different functions. The modular nature of the proposed switch enables that each module can be implemented in any available technology. Thus, with the technology improvements, individual modules can be upgraded as compared to replacing the whole switch. The performance evaluations of the proposed technique show that 3x reduction in processing requirements can be achieved for a 100 channel DWDM-based long haul network. This reduction in processing requirements translates into 2x reduction in hardware cost. Furthermore, the proposed technique provides a large saving in terms of electrical-to-optical and optical-to-electrical converters cost per node. This saving in processing reduction can be leveraged to provide more bandwidth to the services.

In integration of IP and WDM networks, there are two main approaches. One is joint planning approach; the other is joint operation approach. Joint planning typically addresses an optimality problem on total network cost given a certain data traffic pattern and some physical constraints at the network planning stage; while through reconfiguration or real-time setup of lightpaths, joint operation focuses on optimizing network performance or guaranteeing it to a defined level under current network configuration and dynamic data traffic pattern at the network operational stage. On contrary to the (semi-) static virtual topology in joint planning, the joint operation approach maintains a dynamic virtual topology. We propose an integration network model with separated control plane and data plane. Formulations have been studied in building the model, and the initial findings may help in designing the routing and signaling protocols.

This paper studies the performance of a large-scale free-space photonic packet switch based on the generalised knockout principle. Input and output channels are grouped into sectors to make the whole switch a Clos-like three-stage architecture. The input traffic is mainly IP data packets but compatible with ATM cells. Fan-in and fan-out stages are suitable to use free-space optical interconnections with Spatial Light Modulator (SLM) as the switching element. The switch performance was studied based on the switching simulation under different traffic conditions and routing schemes.

We consider wavelength routing networks with and without wavelength converters, and several wavelength allocation policies. We show through numerical and simulation results that the blocking probabilities for the random wavelength allocation and the circuit-switched case provide upper and lower bounds on the blocking probabilities for two wavelength allocation policies that are most likely to be used in practice, namely, most-used and first-fit allocation. Furthermore, we demonstrate that using the most-used or first-fit policies has an effect on call blocking probabilities that is equivalent to employing converters at a number of nodes in a network with the random allocation policy. These results have been obtained for a wide range of loads for both single-path and general mesh topology networks. The main conclusion of our work is that the gains obtained by employing specialized and expensive hardware (namely, wavelength converters) can be realized cost-efficiently by making more intelligent choices in software (name, the wavelength allocation policy).

The passive optical network¹ has been shown to be a cost effective method of introducing fibre to the access network. However, the expected increase in demand for broadband services will inevitably mean that upgrades in capacity will be needed. Wavelength division multiplexing offers an alternative to increasing the bit rate. Two broad strategies exist for the implementation of multiple wavelengths on passive optical networks (PON), fixed wavelength schemes and dynamic wavelength assessment. Wavelength reallocation at call setup allows a flexible response to changes in network demand. It does not, however, take full advantage of statistical multiplexing and the ATM approach. In order to do so, new protocols need to be developed which combine the allocation of wavelengths and time slots. The approach here shows a method based on wavelength allocation at the frame level. A scheme to reduce cell delay variation recently proposed for single wavelength systems is adopted for WDM and an improvement to it demonstrated.

This paper studies the off-line wavelength assignment problem in tree optical networks that use multiple fibers in each link. In the multifiber scenario, the wavelength assignment is potentially more flexible because the same wavelength can be carried simultaneously on parallel links. However, in tree networks, our results unfortunately show that multiple fibers do not fundamentally improve problem complexity nor do they decrease the worst-case bound based on the request load. For multifiber directed or undirected binary trees, we show that the wavelength assignment problem (WAP) is NP-complete. In contract, there is a polynomial algorithm for undirected single-fiber binary trees. If path lengths are restricted to at most three, WAP in dual-fiber undirected tree networks is still NP-complete whereas if path lengths are restricted to four, WAP is any multifiber directed or undirected tree network is NP-complete. Finally, we show a tight upper bound of 3L_max/(2k) wavelengths per fiber for any k-fiber undirected tree where L is the connection load.

Permutation traffic occurs in a number of networking applications. We consider the problem of assigning wavelengths to permutation traffic connections off-line in WDM multi-fiber wavelength-routing ring networks. We focus on a special class of permutation traffic and analyze the bounds on the number of wavelengths required to establish the connections. Lower bounds and simple optimal algorithms are presented.

In this paper, we report on a scalable and reliable switchless wavelength division multiplexing (WDM) network that is based on an arrayed-waveguide grating (AWG). All wavelengths are used for data transmission and signaling is done in-band. Each node at the network periphery is equipped with a single tunable transceiver for data and a broadband light source for control while the network itself is completely passive. Broadcasting is realized by spectrally slicing the broadband signal. The proposed random distributed medium access protocol is reservation based and schedules variably sized data packets on a first-come-first-served and first-fit basis without resulting collisions. The protocol supports both packet and circuit switching and allows for multicasting. The degree of concurrency is significantly increased by using multiple free spectral ranges (FSRs) of the AWG, spatially reusing wavelengths and transmitting data and control informations simultaneously by means of code division multiplexing. Our analytical results demonstrate that exploiting multiple FSRs of an AWG significantly improves the throughput-delay performance of the network.

All-optical networks with wavelength-division multiplexing (WDM) are considered to be a promising technology for next generation transport networks, as they can satisfy the growing bandwidth demand caused primarily due to an explosive growth of web-related services over the Internet. As the traffic demand increases, survivability becomes an indispensable requirement in WDM transport networks. This motivates the need for addressing failure restoration as an integral part of optical network design and operation. To date, the design problems have considered a static traffic demand aimed at optimizing the network capacity and cost, assuming various cost and survivability models. In this paper, we formulate three operational phases viz., initial call setup, medium-term reconfiguration when connections are blocked, and long-term reconfiguration to optimize resource utilization for the existing traffic, as a single Integer Linear Programming (ILP) optimization problem. This integrated framework is an attractive formulation that captures both capacity optimization and service disruption aspect in the problem formulation.

The offset-time-based QoS scheme has been proposed for the next generation Optical Internet as a way to improve current IP's best effort service. For a single node, it has been shown that the offset-time-based scheme efficiently achieves service differentiation without requiring any buffer at the WDM layer. In this paper, the offset-time-based scheme is applied to the multi-hop base. To this end, we consider various policies to handle blocked bursts such as drop, retransmission, deflection routing and buffering in the multi-hop very high performance Backbone Network Service (vBNS), and compare their performance in terms of average wavelength utilization, the average wavelength efficiency and the average end-to-end extra delay. It is shown that the buffering policy is useful with scarce network resource (e.g., bandwidth), but the dropping policy in conjunction with the offset-time-based scheme is good enough with abundant resources.

Transparency of the optical layer offers the possibility to design a network that operates at varying transmission bit rates. While variable bit rate interfaces are being tested and will soon provide the possibility to optimally select the transmission rate for each optical channel, the potential advantages of relying upon multiple transmission rates in the optical network have yet to be fully explored. In this paper we define the concept of Multi-op and Multi-rate (M&M for short) network in which the tributary signal is transmitted over a concatenation of optical channels, with each optical channel operating at its own transmission rate. The optical rate of each optical channel is determined by a number of factors including the end node's interface, amount of multiplexed traffic and cost of the network components. The potential advantages provided by the M&M network when compared to first generation optical networks (i.e., SONET/SDH), to single- and multi-hop (constant bit rate) optical networks, are discussed in general and demonstrated numerically in a WDM ring. Presented results show that the network cost reduction achieved by the M&M design is a function of the cost ratio between the optical bandwidth (wavelengths) and the optical terminals.

To support efficiently the ever growing burst traffic on the Internet (web traffic, broadband multimedia services) over all-optical WDM networks, optical burst switching (OBS) has been proposed as a promising switching method to exploit terabit capacity of optical networks. Fault- tolerant routing for OBS networks has not been studied in the literature. In this paper, we present a distributed scheme of fault- tolerant routing using detours for OBS Networks. Offset time between control packet and subsequent data bursts is varied to accommodate detours. The performance of the routing scheme is evaluated, through simulation, in terms of the burst loss probability for different network parameters like a number of wavelengths and the traffic load. We show that by tuning the offset time, the routing can very efficiently achieve low burst loss probability.

We consider a packet-switched network overlaid on a wavelength-routing optical network. A pre-designed path protection scheme to restore traffic upon a single link or node failure in bidirectional rings is presented. An approach to restore affected traffic using a pre-designed logical topology that is embedded after failure on the surviving topology is discussed. We present several logical topology designs with the goal of minimizing the maximum logical distance between any two nodes. The designs are rigorously analyzed and bounds (which are tight in some cases) are presented.

In this paper we propose a new approach for optical layer restoration management in a dynamically reconfigurable optical network. The optical network consists of Optical Layer Cross-connects (OLXCs) which have the ability to convert the wavelength of any incoming channel to any outgoing wavelength (i.e. have wavelength conversion). Recently, there has been intense interest within the optical networking community and in standards bodies on bandwidth management in a dynamically reconfigurable optical networks [e.g. 1,2,3,4]. In these networks, it is assumed that the physical hardware is deployed, but that network connectivity is not defied until high bandwidth connections referred to as linghtpaths are established within the network. The lightpaths are provisioned by choosing a route through the network with sufficient available capacity. The lightpaths are established by allocating capacity on each link along the chose route, and appropriately configuring the OLXCs. Restoration is provided by reserving capacity on routes that are physically diverse to the primary lightpath. In this paper we present a new approach to restoration for the future optical Internet, employing distributed restoration bandwidth measurement, failure detection and failure handling.

We propose a programmable automatic protection switching (APS) protocol to repair an impaired lightpath traversing an optical link. Recovery agents repair impaired flows by searching through a space of policies before attempting a protection switch and after switching impaired traffic. A policy manager disseminates a changeable set of policies to each agent and ensures consistent interpretation end-to-end QoS. QoS policies are structured to be interpreted in the same way by developing a model of end-to-end QoS over which logic formulae can be checked for satisfaction.

In this paper, we evaluate the performance of several multicast schemes in optical burst-switched WDM networks taking into accounts the overheads due to control packets and guard bands (Gbs) of bursts on separate channels (wavelengths). A straightforward scheme is called Separate Multicasting (S-MCAST) where each source node constructs separate bursts for its multicast (per each multicast session) and unicast traffic. To reduce the overhead due to Gbs (and control packets), one may piggyback the multicast traffic in bursts containing unicast traffic using a scheme called Multiple Unicasting (M-UCAST). The third scheme is called Tree-Shared Multicasting (TS-MCAST) wehreby multicast traffic belonging to multiple multicast sesions can be mixed together in a burst, which is delivered via a shared multicast tree. In [1], we have evaluated several multicast schemes with static sessions at the flow level. In this paper, we perform a simple analysis for the multicast schemes and evaluate the performance of three multicast schemes, focusing on the case with dynamic sessions in terms of the link utilization, bandwidth consumption, blocking (loss) probability, goodput and the processing loads.

In this paper, we propose a novel scheduling scheme, namely Packet Preemptive Scheduling (PPS), which is designed to handle real-time (high priority) packets in WDM star networks. PPS allows high priority packets to displace the scheduled low priority (non-real-time) packets. By doing so, PPS guarantees that the high priority packets can always achieve the earlier transmission than the others so as to meet the QoS requirements. This scheme has the capability of handling channel collision and destination conflict. The performance of the PPS scheme has been extensively studied by means of numerical simulations.

Limited signalling capacity imposes severe constraints on the control protocol flexibility of a national-scale WDMA/TDMA switchless optical transport network architecture. To signal the maximum number of connection requests for optical packets from the terminals within each transmission frame, out-of-band signalling in dedicated signalling wavelength channels would require at least 12 receivers in each terminal; an excessively expensive proposition. In-band meta-signalling techniques, using all of the network's available wavelength channels, together with temporal manipulation of blocks of wavelength channels by fibre delay line, ensure that the packet connectivity of a future WDMA/TDMA switchless optical transport network is completely unconstrained by any signallying limitations, at minimum cost.

IP routers can be coupled with wavelength-selective optical cross- connects to support existing Internet infrastructure in a wavelength division multiplexing (WDM) optical network. Because optical wavelength routing is transparent to IP, packets can bypass traditional forwarding and pass directly through the optical cross-connect, resulting in very high throughput and low delay routing. This approach shares features with label switching, but wavelengths are much more scarce resource than labels. Because optical switches have larger switching times than electronic switches, and wavelength conversions are expensive, wavelength label swapping is not easily done. Wavelength label assignments must consider these limitations to be practical in an optical environment. The performance of an instance of this approach, called Packet over Wavelengths (POW) has been simulated and studied. A new signaling protocol, Simple Wavelength Assignment Protocol (SWAP) is devised to be POW signaling protocol. SWAP takes into account the optical device limitations, and is designed to minimize wavelength conversion, utilize wavelengths with the merging of flows, and reduce the reconfiguration of optical switches. SWAP, to our knowledge, is the first approach to combine signaling and wavelength assignment in an on- line protocol. This paper describes high level SWAP design challenges, decision, and overhead.

This paper presents a distributed control protocol for wavelength reservation in wide-area WDM all-optical networks. The protocol aims to establish wavelength-continuous lightpaths dynamically and efficiently so as to minimize the overall blocking probability at the cost of a nominal increase in control overhead. It uses forward reservation with immediate unlocking of unnecessary wavelengths at intermediate nodes to reduce the change of blocking. Forward reservation is normally considered to be better than backward reservation from the point of view of set-up delay. Immediate unlock facility makes it more efficient in terms of blocking probability. We show through simulation that the protocol improves the success rate in connection establishment (considerably for low aggressiveness) without much incrementing the control overhead.

Given the significant progress made and the continuing advances expected in the optical networking technology, it becomes attractive to build a future Optical Internet that natively supports bursty IP datagrams. Burst switching WDM optical networks are touted as suitable network architectures for future Optical Internet backbones. However, the lack of optical processing capabilities results in increased burst blocking probability, which in turn leads to very limited network performance. Efficient contention resolution method is therefore necessary. Based on discussions of the state of the art of recent optical technologies, a deflection routing protocol for burst switching WDM mesh networks is proposed. The idea of this approach is to use idle optical links as fiber delay lines for contention resolution. Simulation results show that the proposed protocol is available solution for effectively reducing the blocking probability and increasing the performance of burst switching WDM optical networks.

In this paper, we analyze the performance of WDM networks with traffic grooming capabilities supporting low-rate circuit-switched traffic streams. Traffic grooming in WDM networks collectively refers to the multiplexing, demultiplexing and switching of lower-rate traffic streams onto high capacity lightpaths. Networks which perform grooming only at the OADMs present in the nodes are referred to as Constrained Grooming Networks. Networks whose nodes switch traffic streams between wavelengths and perform grooming at the OADMs are referred to as Sparse Grooming Networks. Given the network topology, the traffic matrix and the node locations of grooming and traffic stream switching, we present an analytical model, using link-independence and wavelength-independence assumptions, to calculate the blocking performance. We illustrate the benefits of sparse grooming over constrained grooming in the mesh-torus and ring network topologies, using both simulation and analytical results.

In this paper, we propose three methods of improving the performance of all-optical DWDM networks that employ light-path reservation. They are wavelength conversion, multiple fibers and two-path routing. We being by developing analytical models for static-routing networks based on the bidirectional Manhattan Street network (BMSN) and de Bruijn (DB) network topologies. Analytical results indicate that the multiple fibers approach yields the lowest blocking probability for BMSN topologies. In addition, one can achieve the greatest improvement in de Bruijn networks using either wavelength conversion or multiple fibers. The choice depends on the values of the parameters such as network load and reservation level. The two-path routing method offers moderate improvement in perfomrrance for the de Bruijn network topology. In addition, there exist optimum reservation levels for attaining minimum blocking probabilities. In the de Bruijn network, the advantage of the three schemes is modest. However, these schemes offer significant performance improvement for the bidirectional Manhattan Street network.

The public network is rapidly moving from being a voice-optimized network that carries data, to a data-optimized network that will carry multiple services with IP as the predominant traffic. One of the key issues confronting carriers is the balance between circuit switching and packet switching in the core of the network. With the emergence of fast IP switch/routers, it is becoming possible to directly connect all ports on DWDM equipment directly on the router vs. connecting the router to DWDM infrastructure through an optical cross-connect. Comparing the costs of circuit switching and packet switching in the core of the network, a hybrid of packet and circuit switching solution proves more cost-effective than the pure packet switching whenever the port cost of the IP router is more than twice the port cost of the optical cross- connect. Furthermore, to enable cost effective evolution toward the converged network, the next generation of switching equipment will need to accommodate astronomical growth, while providing carriers with more flexibility in network design.

The access network plays a major role in meeting the rapidly growing demand for higher bandwidth by transporting traffic from the long-haul network to the end users. The distribution network is part of access network that carries traffic to and from the customer premises to the feeder network. Distribution networks are usually passive with short span length. One of the major impairments in a distribution network is crosstalk which represents leaking of signal from one channel to another. Crosstalk along with various types ofnoise at the receiver degrades the quality of signal. This paper presents a qualitative and quantitative analysis of the effects of crosstalk on signal quality and system performance. The output power, worst case crosstalk, and system performance in terms of Q value are calculated for different device parameters and network configuration to evaluate the impact of crosstalk on the quality of signal and system.