Wavelength utilization in WDM wavelength routing networks can be improved by time division multiplexing sessions on each wavelength, and switching time slots and wavelengths. In this paper, we consider the problem of assigning time slots and wavelengths to a given set of multi-rate sessions in WDM/TDM buses and rings. Given a set of sessions and their time slot requirements, our objective is to maximize the network throughput. Bounds on frame length, and hence network throughput, under two different scheduling constraints are presented. Simulation results using a simple assignment algorithm that achieves near-optimal frame lengths are also presented.

We consider a broadcast-and-select (single hop) photonic ATM switch architecture which operates under schedules that mask the transceiver tuning latency. Within this work, we develop a set of expressions to obtain the probability distribution of a pair of cells which arrive to an WDM switch consecutively being separated upon retransmission out of the switch by a period of t slots. This measure warrants investigation in that it is a necessary first step to characterizing the impact of a WDM architecture on the delay jitter of a data flow. The analysis is carried out using arrival models that capture the notions of burstiness and correlation, two important characteristics of ATM traffic. The results show that the impact of the switch architecture on the delay variation of two cells' transversing the switch may be significant and unintuitive, but can be predicted based on scheduling parameters and other traffic patterns.

Recent developments in optical WDM technology have opened up many new challenges and opportunities for network designers. The availability of increasingly advanced optical devices, such as add-drop-multiplexers and wavelength routers/cross- connects, is pushing this technology beyond simple point-to- point deployment and more towards applications in end-to-end networking. These trends will require more advanced network control features in order to take full advantage of the optical fiber bandwidth. Although much work has been done on static WDM network provisioning, the issue of lightpath channel control in practical network environments needs further investigation. New protocols are required to provision channels for various higher-layer protocols, and these schemes must make efficient use of key network resources such as wavelengths and switching/conversion capacity. Optical channel survivability is also an important concern and various types of schemes can be envisioned.

HLAN is a promising architecture to realize Tb/s access networks based on ultra-fast optical TDM technologies. This paper presents new research results on efficient algorithms for the support of quality of service over the HLAN network architecture. In particular, we propose a new scheduling algorithm that emulates fair queuing in a distributed manner for bandwidth allocation purpose. The proposed scheduler collects information on the queue of each host on the network and then instructs each host how much data to send. Our new scheduling algorithm ensures full bandwidth utilization, while guaranteeing fairness among all hosts.

This paper presents a fast approach to interactively designing optimal resilient WDM mesh networks based on an optimization engine that, given the network layout, the light path demands and the cost function, sub-optimally selects a ring cover and, for each lightpath demand, the route and the protecting rings. The three selections are made jointly as it has been demonstrated that a 15% cost reduction can be achieved in this way when compared to a sequential selection. The optimization engine, based on Simulated Annealing, yields satisfactory solutions with CPU times ranging from seconds to few minutes, thus allowing user interactive design of the system. In addition, the inherent flexibility of SA allows us to minimize the system cost under various Multi-WSHR protection schemes, including Dedicated-Path-switched WSHR and Shared-Line-switched WSHR. A last feature of the proposed approach is the capability to handle network designs in which, due to particular network layout and/or technology constraints, not all the lines of the mesh can be protected. Under these circumstances, the lightpath demands routed on the unprotected lines are minimized.

Wavelength-routed all-optical networks have been receiving significant attention for high-capacity transport applications. A good routing and wavelength assignment algorithm is critically important to improve the performance of wavelength-routed WDM networks. We study the blocking performance of fixed-paths least-congestion (FPLC) routing in multifiber WDM networks in this paper. A new analytical model based on the link-load correlation is developed to evaluate the blocking performance of the FPLC routing. The analytical model is a generalized model that can be used in both regular (e.g. mesh-torus) and irregular (e.g. NSFnet) networks. It is shown that the analytical results closely match the simulation results, which indicates that the model is adequate in analytically predicting the performance of the FPLC routing in different networks.

Fault management in WDM routed all-optical networks has mostly been addressed either by automatic protection switching or through loop-back recovery. These schemes are designed for managing single fault occurrence and generalization method to handle multiple faults are not known. Conventional routing schemes are static in nature (where the routers are programmed to realize the lightpaths between the end-nodes) and hence a fault management scheme needs to find a fault-free path between end-nodes using the settings of the routers. This paper considers the principle of survival route graphs to construct fault-free paths between end-nodes. As a result, the fault avoiding route between two end-nodes might be a multihop route in which the number of hops are limited to reduce the communication delay. The performance degradations of the network because of fault occurrence are studied through simulations and measured in terms of blocking probability and communication delay.

This paper addresses the issues of guaranteed and scalable end-to-end QoS in Metropolitan DWDM networks serving as transit networks for IP access networks. DWDM offering few wavelengths have in the past been deployed in backbone networks to upgrade point-to-point transmission where sharing is based on coarse granularity. This type of DWDM backbone networks, offering few lightpaths, provides no support for QoS services traversing the network. As DWDM networks with larger numbers of wavelengths penetrate the data-centric Metro environment, specific IP service requirements such as priority restoration, scalability, dynamic provisioning of capacity and routes, and support for coarse-grain QoS capabilities will have to be addressed in the optical domain in order to support end-to-end Service- Level Agreements. In this paper, we focus on the support of QoS in the optical domain in order to achieve end-to-end QoS over a DWDM network. We propose a QoS service model in the optical domain called Differentiated Optical Services (DOS). Service classification in DOS is based on a set of optical parameters that captures the quality and reliability of the optical lightpath.

A novel architecture is proposed for future multi-terabit IP (internet protocol) routers, employing multiple cascaded stages of optical switching and buffering. WDM is used within the node to facilitate its operation. External synchronization is not required, and a void-filling algorithm is used to simplify hardware requirements. Packet priorities are not implemented in the current version of the switch, and the issue of header table lookup is not considered. Performance with respect to packet loss is studied by simulation, demonstrating that this multi-stage concept results in substantial hardware reduction.

As optical device technology progresses, it will become feasible to implement more complex switching nodes that can perform Individual Switching of the Wavelengths within a time slot (IWS). While IWS nodes can be expected, at least in the beginning, to be more costly than Photonic Slot Routing (PSR) nodes, they offer increased switching flexibility, thereby potentially improving the network performance. This paper presents a solution that supports a gradual migration from a PSR-based network to an IWS-based network, by replacing selected PSR nodes with IWS nodes.

Optical burst switching (OBS) is a promising solution for building terabit optical routers and realizing IP over D- WDM. In this paper, we describe the basic concept of OBS and a general architecture of optical routers. The key design issues related to the OBS are discussed, notably, burstification, offset-time management and channel scheduling. A simple channel scheduling algorithm called LAUC (Latest Available Unscheduled Channel) algorithm is presented, which is also called Horizon algorithm in the literature. The performance of optical routers under random and self-similar traffic is thoroughly studied via computer simulations. Our study indicates that the distribution of burst length is not exponential any more. For the FDL (fiber delay line) optical buffer, the burst loss ratio is quite sensitive to the traffic characteristic of bursts when using the LAUC algorithm.

We have used optical simulations as a means of setting component requirements, assessing component compatibility, and designing experiments in the MONET (Multiwavelength Optical Networking) Project. This paper reviews the simulation method, gives some examples of the types of simulations that have been performed, and discusses the validation of the simulations.

SIMON is an object-oriented event-driven simulation package implemented in C++ which incorporates optical device characteristics in the measurement of network-level blocking statistics. SIMON is suitable for studying the performance of large wavelength-routed optical networks, in which a call is set up in the network for a specific duration on a pre- determined lightpath. Currently the physical-layer models allow for modeling phenomena such as signal attenuation in fiber and other components, amplifier gain saturation, and homowavelength crosstalk in switches. Simulation experiments can be performed with a user-specified bit-error rate limit, which must be satisfied by any call set up in the network.

We present here the innovative application of Artifex modeling and simulation environment for the design of optical networks and protocols. Artifex is used to design discrete-event systems, with a visual representation of dynamics and very efficient discrete-event simulation and software generation. Artifex is a very stable and user- friendly environment that has been used, for example, in the design and simulation of traditional routers and switching network architectures.

CATO (prototype CAD Tool for Optical Networks and Interconnects) performs DWDM network design, balancing costs and performance. CATO's interactive visual interface allows demand conversion; topology creation/editing; protection; lightpath routing and wavelength assignment; reduction and placement of OXC ports, (lambda) -converters, amplifiers, and other critical resources; network simulation; and graphical reporting. CATO combines novel heuristics with artificial intelligence, solving problems unique to optical networks, yielding dramatic savings, and running in half the time or less of tools based on linear programming. Thus, CATO can maximize performance, utilization, and survivability in real time, and is suited for management tasks.

Earlier efforts on optical access concentrated on the design of PONs for the collection and distribution portion of the access network. In these networks the optical hardware in the RN is very simple, but a multiple access protocol is needed for upstream traffic control. On the other hand, the role of communications is already well established in the office environment. With the advent of cheap, affordable broadband communications and the increasing complexity of consumer electronics, it seems natural to extend the network into the home. As the application of Home Area Network is ever increasing, we therefore consider connectivity between access network and home network which generates various traffic to design MAC protocol over residential network. Global-FIFO is quite simple and allows dynamic upstream bandwidth assignment on the basis of a request-and-permit mechanism. It has good bandwidth efficiency and being cell- based, it does not consider the various traffic from home network. In this paper, we design and analyze the new MAC resource assignment algorithm called MQ-FIFO (Multiple Queue-FIFO) that provides good performance under the environment of ATM-PON and Home Network.

A wavelength division multiplexed network is considered for arbitrary topologies. The network allows optical signals to pass through nodes, which often results in less electronic and opto-electronic equipment than networks that do not, i.e., networks that only switch traffic electronically. A disadvantage of having optical pass through is that there is less capability to switch tributary traffic streams and so there may be more blocking. However, it is shown that the network operates as well as a network with only electronic switching under a particular incremental traffic model. Examples are given that shows that the network can have lower cost when the cost is dominated by the line terminating equipment. A simple heuristic design algorithm is also given to configure the network to minimize its cost.

The introduction of wavelength division multiplexing into the access network is a potentially attractive alternative upgrade strategy to increasing the bit rate of passive optical networks (PON). A number of different methods have been proposed to implement a multi-wavelength PON and hence a careful assignment needs to be made of the most cost effective, both in the short term and for future increases in capacity. One key issue is the use of dynamic allocation of wavelengths, which allows the reconfiguration of the network in response to network demand. A model has been produced to simulate demand patterns on a PON, in order to quantify the benefits of dynamic assignment.

The next generation of optical communication networks should provide a significant role for large range high-splitting ratio high-speed passive optical networks (SuperPONs). Optical amplification of both upstream and downstream traffic is necessary to sustain these configurations with several amplifiers in series throughout the network. The Photonic Local Access Network (PLANET), an EU-ACTS project, was developed to operate a 2 - 5 Gbit/s in the downstream direction (from the head-end to the user optical network unit) and at 311 Mbit/s in the upstream. A laboratory-based system demonstrated the feasibility of the SuperPON concept over a span of 100 km to support a potential 2,048 optical network units. The downstream utilized erbium-doped fiber amplifiers whereas in the upstream the use of burst-mode semiconductor optical amplifiers minimized noise funneling effects. Simulation results are presented in this paper for the upstream SOA cascade. Data are provided for the evolution of both the signal power and the amplified spontaneous emission power throughout the amplifier cascade.

This paper describes the TDMA uplink of the ACTS PLANET project, realizing a high split long range access network using optical amplification. An overview of the network topology and the different building blocks is given. In particular, the differentiating burst mode receiver, the optical repeater units and the amplified splitter controller will be discussed. Bit error measurements were performed to define the system limits of the access network. It is shown that a feeder length of 90 km, in combination with a split factor of 2000 and a drop section length of 10 km are feasible.

Currently, the first FSAN compliant ATM-based Passive Optical Networks (APON) are installed in the field. A possible evolution scenario for these types of access networks could be the so-called SuperPON system. The SuperPON system exploits all possible upgrades of an FSAN APON system. It comprises namely an upgrade on the following parameters: range: 100 km instead of 20 km; split: 2048 instead of 64; and bitrates: 2.4 Gbit/s downstream instead of 622 Mbit/s and 311 Mbit/s upstream instead of 155 Mbit/s. Due to these significant changes in system parameters, it becomes of the utmost importance to implement a dynamic Medium Access Control (MAC) protocol on the SuperPON, since the upstream bit rate is rather limited for such a large number of connected Optical Network Units. Main challenges specific to the SuperPON are the impact of long range and high number of users on the efficiency of the MAC protocol.

In this paper, we present an analytical model to analyze and to evaluate the blocking performance of a WDM-based All- Optical Network when light-path reservation is introduced. The computed results show that the link blocking probability is marginally less with light-path reservation than when there is no reservation provided that the traffic is low and that no routing is allowed; while the path blocking probability reduces significantly at high traffic when light-path reservation is introduced. Furthermore, the mean blocking probability of the network is reduced for low traffic with light-path reservation and routing is allowed. However, for high traffic, light-path reservation increases the blocking probability. We also discovered that, for given traffic load, there exists an optimum reservation level at which one can achieve the minimum mean blocking probability.

Recently, the number of carriers have increased and they intensively compete with each other. In order to accommodate skyrocketing traffic, it is natural that several carriers newly construct fiber optic networks as well as the existing ones. The wavelength division multiplexing (WDM) technology is promising for the construction. In the WDM networks, several routing and wavelength assignment strategies have been proposed and the cost effectiveness have been shown in sample topologies. However, such topologies will not always meet the future's great traffic demand. In this paper, we propose a new topological design algorithm to minimize network cost consisting of the conduit and fiber optic cost. We validate the minimization level of our heuristic algorithm by comparing it with lower bound.

Dynamic routing in WDM (Wavelength Division Multiplexing) networks is achieving increasing interest due to the expected traffic growth for future optical networks. This contribution concentrates on network performance issues occurring for network operation under dynamic traffic load. In this context, routing strategies play a central role. We show that a new strategy achieves very good performance over a wide load range by making use of three beneficial components: providing multiple pre-calculated alternatives which are selected according to the network planning decisions, dynamic path search to search for routes in the cases where all pre-calculated alternatives are blocked, and an adaptive length limit to avoid on the one hand unnecessary restrictions for low traffic load, and on the other hand very long alternative routes for high traffic load. Moreover, specific effects for the influence of resource allocation strategies in photonic WDM networks are highlighted. This includes the dependence of a routing strategy on conversion degree and the influence of converter usage strategies in networks with partial conversion. The results show that an efficient routing strategy has to take into account various specific aspects of WDM networks to achieve good performance. Finally, we present how results from dynamic routing investigation can help to optimize the network planning process.

In this paper we have examined the efficacy of virtual wavelength conversion in all-optical networks with respect to full wavelength conversion. It is well known that wavelength conversion reduces the blocking probability. All- optical full wavelength converters are not yet commercially available and wavelength conversion typically involves opto- electrical conversion. Virtual wavelength conversion avoids any change in the carrier frequency and is feasible using available technology. In this paper we have studied a dynamic wavelength assignment scheme using a distributed algorithm. We have shown that using virtual wavelength conversion, we can achieve the effect of full wavelength conversion without using wavelength converters in optical domain.

This paper considers a network design problem introduced which resource allocation is performed within a Wavelength Division Multiplexing optical ring network to reduce overall equipment cost. The problem can be summarized as follows. The Add-Drop Multiplexers (ADM's) that are used to terminate lightpaths in an optical (SONET/SDH) ring are expensive and therefore a naive implementation that places ADM's for each wavelength at every node would waste expensive resources. However, by carefully assigning connections to wavelengths and routing the resulting lightpaths, one may avoid terminating all wavelengths at every node, and thereby reduce the number of ADM's needed.

Using a novel and simple Networking Planning and Optimization Tool, it is shown that a full-mesh connected 4- fiber WDM ring network is capable of supporting up to 15- nodes employing 28-wavelength channels at 10 Gb/s/channel, while simultaneously maintaining both smooth network evolution and adequate end-to-end performance, independent of the network size, for all wavelength channels across the network.

We present new theoretical results on the performance limits of on-line routing and wavelength assignment (RWA) algorithms in a single-hub wavelength division multiplexing ring network architecture. An RWA algorithm is said to be on-line if at each point in time, the algorithm assigns a lightpath to a current connection request based only on past information and with no knowledge whatsoever about the future requests. We derive tight bounds on the throughput performance of an on-line min-hop algorithm in comparison with that of an optimal off-line algorithm. In addition, we show that when there are sufficiently many nodes in the ring, the min-hop algorithm achieves the best possible throughput performance among all on-line RWA algorithms. On the other hand, when the number of nodes is not greater than the number of wavelengths in the ring, an on-line algorithm that minimizes the maximum load of the network is shown to achieve better throughput performance than the min-hop algorithm.