The presentation will introduce the past and ongoing research and development activities of the Department of Telecommunications of Budapest University of Technology and Economics (BUTE). The workgroups of the Department focus on a number of different research topics, such as modeling, design and analysis of telecommunications networks, modeling, measurement and analysis of network traffic, and other related problems. The presentation will give an overview of the various international and national projects related to the research activities of the Department.

This talk presents a novel protection strategy that copes with network multiple consecutive and unrelated failures in the presence of dynamic traffic. The strategy is based on the re-routing of orphan end-to-end connections upon occurrence of the (first) failure.

In reliable Wavelength Division Multiplexed (WDM) networks the optimal choice of Routing and Wavelength Assignment (RWA) for the working and protection path-pair is often a complex problem to solve. To cope with such problem complexity, the Disjoint Path-Pair Matrix (DPM) approach has been introduced and studied. With the DPM approach the RWA problem complexity can be significantly reduced by limiting the number of candidate path-pairs considered in the optimization. Simulation results are collected using the DPM approach to solve the online RWA problem in a network based on the Shared Path Protection switching scheme with Differentiated Reliability. When compared to the conventional k-shortest paths approach, the DPM approach requires up to one order of magnitude less candidate path-pairs. In addition, the DPM approach finds solutions with reduced hop length of both the working and protection paths by up to 3% and 14%, respectively.

The capacity of a Wavelength Division Multiplexing (WDM) network can be improved by a selective upgrade of the optical processing capabilities of its wavelength/waveband routing nodes. In the case of wavelength routing, the focus is on optimal placement of wavelength converters and on optimal tuning of wavelength changing facility. The investigation has been carried out in a wavelength converting network with regular mesh and ring topologies under static and dynamic traffic scenario. An efficient criterion is proposed to trade off between the blocking rate and the number of nodes that allow wavelength conversion and their percentage of wavelength conversion facility. The network average link usage is monitored to ensure proper utilization of the network resources, to validate the optimization design and to draw conclusions about the network scalability.

The advantage of hybrid OXCs which involve both all optical (OOO) and electrical(OEO) switching lies in the reduction of swtiching cost by aggregating wavelength paths into waveband paths and switching them through the transparent optical layer. Simulation studies have been carried out in a joint wavelength and waveband routed networks to select the wavelength convergence point(WCP) and the wavelength divergence point(WDP) nodes on the call blocking probability for a given static traffic.

Wavelength routing networks resemble previous circuit-switching networks, as the lightpath established between two nodes may be considered as a circuit between them. Although there has been recently much interest in other ways of switching in all-optical networks (ex.: packet and burst switching), circuit (or lightpath) switching is still the most promising approach today for optical networks. Among the performance metrics for such networks, blocking has been the most important one in its design. Then, given the existence of previous theory on circuit-switching blocking networks, there was a natural will to reuse its results on the blocking performance of all-optical networks, focusing on the their peculiarities: the wavelength continuity constraint when nodal conversion capabilities are not available; and the relatively poor connectivity of the first optical networks. Most of former discussions were based on the Lee approximation, which is particularly inaccurate when nodal degrees are small, as in current optical networks.
The objective of this talk is to discuss some results of existing theory of blocking in wavelength routing networks and to present new models of blocking that have been investigated by our group recently. We propose a better substitute for the independent link assumption present in the classical Lee approximation. The classical independent link assumption recognizes the existence of only two kinds of objects in the network: single free (available) links, which do support requests for new paths; and single busy links, which donot support them. In real networks, however, independent single busy links may occur only in 1-link paths. Any path with H > 1 links will give rise to a set of H spatially connected busy links. The new assumption then takes all active paths, as well as all free links, as independent objects on the network topology. The new model is shown to generate a very good fit with the blocking probabilities obtained through simulations on linear topologies using Poissonian, spatially homogeneous traffic. We also discuss the independent wavelength assumption in our analysis. Some results of our theory will be published this year in SPIE Opticomm 2003 and IEEE Globecom 2003.

The objective of the presentation is to compare three well-known WDM network architectures (first-generation, single-hop, and multi-hop) when they are deployed to accommodate dynamic end-to-end connections that require sub-wavelength transmission rates. The comparison is based on a performance figure that is uniquely defined to take into account the various architecture costs -- determined by the cost of the deployed network elements. The defined performance figure permits also to compare the three architectures for all possible line-to-node cost ratio values.

The Differentiated Reliability concept is developed by a in OPNEAR lab with Centralized Connection Management. The concept is extended to be applied to Distributed Connection Management. The initial part of the study focuses on Optical Mesh Network for Emerging Gigabit Application (OMEGA) Test bed findings.

In order to compare the concept with pure Shared Path Protection scheme for larger topology, Network Simulator is used. Connection blocking in 19 node network is analyzed. When resource reservation for a connection fails, either due to lack of resources in the network or due to contention of a resource as a result of distributed control; it is not dropped from the network. The connection is reattempted until it is successful. Effects of reattempt strategy in case of First Fit (FF) and Random Fit (RF) of routing and wavelength assignment (RWA) are studied. A connection will be dropped from the network only when the source node server is busy serving a previous connection.

Highly time-sensitive network applications require fast recovery not only from failure of single network element but also from simultaneous failures of multiple network elements. Current local recovery schemes in MPLS TE networks achieve 10s of ms recovery time under the assumption of single network failure. This paper is aimed to extend the applicability of MPLS local recovery to multiple failure scenarios. The proposed approach classifies multiple failure scenarios into different risk levels based on the fact that each scenario features different failure probability. Multiple failure scenarios increase substantially the difficulty both in finding the bypass tunnels around the failure and in identifying the failure scenario. Two approaches, fault-dependent approach and fault-independent approach are proposed by taking the risk level information into consideration during both bypass tunnel computation and selection. These two approaches trade recovery time with recovery ratio.