András Faragó

Lightpath (wavelength) routing in large WDM networks

We address the problem of efficient circuit switching in wide area optical networks. The solution provided is based on finding optimal routes for lightpaths and the new concept of semilightpaths. A lightpath is a fully optical transmission path, while a semilightpath is a transmission path constructed by chaining together several lightpaths, using wavelength conversion at their junctions. A fast and practical algorithm is presented to optimally route lightpaths and semilightpaths taking into account both the cost of using the wavelengths on links and the cost of wavelength conversion. We prove that the running time of the algorithm is the best possible in the wide class of algorithms allowing linear algebraic operations on weights. This class encompasses all known related practical methods. Additionally, our method works for any physical realization of wavelength conversion, independently whether it is done via optoelectronic conversion or in a fully optical way.

Optimizing the system of virtual paths

The virtual path (VP) concept is known to be a powerful transport mechanism for ATM networks. This paper deals with the optimization of the virtual paths system from a bandwidth utilization perspective. While previous research on VP management has basically assumed that bandwidth in ATM networks is unlimited, emerging technologies and applications are changing this premise. In many networks, such as wireless, bandwidth is always at a premium. In wired networks, with increasing user access speeds, less than a dozen of broadband connections can saturate even a Gigabit link. We present an efficient algorithm that finds a system of VP routes for a given set of VP terminators and VP capacity demands. This solution is motivated by the need to minimize the load, or reduce congestion, generated by the VPs on individual links. A nontrivial performance guarantee is proven for the quality of the proposed solution and numerical results show that the proposed solution carries the potential for a near optimal allocation of VPs. >

Fast nearest-neighbor search in dissimilarity spaces

A fast nearest-neighbor algorithm is presented. It works in general spaces in which the known cell techniques cannot be implemented for various reasons, such as the absence of coordinate structure or high dimensionality. The central idea has already appeared several times in the literature with extensive computer simulation results. An exact probabilistic analysis of this family of algorithms that proves its O(1) asymptotic average complexity measured in the number of dissimilarity calculations is presented. >

Shared path protection with differentiated reliability

The authors (Fumagalli and Tacca (2001)) introduced the concept of differentiated reliability (DiR) applied to dedicated path protection (DPP) switching in wavelength division multiplexing (WDM) rings. By means of the DiR concept, a network can be designed to provide multiple degrees of reliability and efficiently satisfy the user-specific requirements, yet minimizing the network total cost. This paper extends the DiR concept to the case of shared path protection (SPP) switching in arbitrary (mesh) topology, the so called SPP-DiR. A time efficient algorithm is proposed to determine the primary and backup path of each demand in both conventional SPP and SPP-DiR (WDM) networks. When compared to DPP, results obtained for the pan-European network by using the proposed algorithm indicate cost reductions of about 16% when SPP is applied, and up to 34% when SPP-DiR is applied.

A new degree of freedom in ATM network dimensioning: optimizing the logical configuration

A mathematical model is presented that provides a well-defined formulation of the logical configuration problem of ATM networks (the carriers of future B-ISDN) with the objective of maximizing the total expected network revenue, given the physical network parameters and the traffic requirements of each virtual subnetwork. A two-phase solution procedure is developed in which the decision variables are the logical link capacities that specify the logical decomposition into virtual subnetworks, and the load sharing parameters. The first phase of the solution finds a global optimum in a rougher model. The second phase uses this as an initial point for a gradient-based hill climbing that applies the partial derivatives of the network revenue function obtained in a more refined model. >

Strong universal consistency of neural network classifiers

In statistical pattern recognition, a classifier is called universally consistent if its error probability converges to the Bayes-risk as the size of the training data grows for all possible distributions of the random variable pair of the observation vector and its class. It is proven that if a one-layered neural network with properly chosen number of nodes is trained to minimize the empirical risk on the training data, then a universally consistent classifier results. It is shown that the exponent in the rate of convergence does not depend on the dimension if certain smoothness conditions on the distribution are satisfied. That is, this class of universally consistent classifiers does not suffer from the curse of dimensionality. A training algorithm is presented that finds the optimal set of parameters in polynomial time if the number of nodes and the space dimension is fixed and the amount of training data grows. >

A deterministic approach to the end-to-end analysis of packet flows in connection-oriented networks

We analyze the worst-case behavior of general connection-oriented networks, with first-in-first-out (FIFO) queueing policy, forwarding packets along an arbitrary system of routes. A worst-case bound is proven for the end-to-end queueing delay and buffer size needed to guarantee loss-free packet delivery, given that sources satisfy a given source rate condition. The results are based on a novel deterministic approach and help in reconciling the discrepancy between the unstable worst-case behavior of FIFO-based networks and their good practical performance.

On the stability of paths, Steiner trees and connected dominating sets in mobile ad hoc networks

We propose algorithms that use the complete knowledge of future topology changes to set up benchmarks for the minimum number of times a communication structure (like paths, trees, connected dominating sets, etc.) should change in the presence of a dynamically changing topology. We first present an efficient algorithm called OptPathTrans that operates on a simple greedy principle: whenever a new source-destination (s-d) path is required at time instant t, choose the longest-living s-d path from time t. The above strategy when repeated over the duration of the s-d session yields a sequence of long-lived stable paths such that number of path transitions is the global minimum. We then propose algorithms to determine the sequence of stable Steiner trees and the sequence of stable connected dominating sets to illustrate that the principle behind OptPathTrans is very general and can be used to find the stable sequence of any communication structure as long as there is a heuristic or algorithm to determine that particular communication structure in a given network graph. We study the performance of the three algorithms in the presence of complete knowledge of future topology changes as well as using models that predict the future locations of nodes. Performance results indicate that the stability of the communication structures could be considerably improved by making use of the knowledge about locations of nodes in the near future.

ADAPT: a dynamically self-adjusting media access control protocol for ad hoc-networks

This paper presents a dynamically adaptive protocol for transmission (ADAPT) for ad hoc networks that combines, in a novel way, a collision-free allocation based protocol and a contention based protocol while retaining the advantages of each. At low loads, ADAPT uses its contention mechanism to reclaim/reuse bandwidth that would otherwise be wasted by a pure allocation based protocol. At high loads, ADAPT provides bounded delay guarantees by dynamically changing its operation to that of its allocation based protocol, avoiding the fundamental problem of instability associated with pure contention based protocols. Thus, ADAPT self-adjusts its behavior according to the prevailing network conditions. Both analysis and simulation results demonstrate that the two protocols interact in a positive way, showing that it is possible to combine the advantages of two fundamentally different design philosophies without suffering from their drawbacks.

Meta-MAC protocols: automatic combination of MAC protocols to optimize performance for unknown conditions

A systematic and automatic method to dynamically combine any set of existing MAC protocols into a single higher layer, or meta-MAC protocol, is presented. The new approach makes it possible to always achieve the performance of the best component protocol, without knowing in advance which protocol will match the potentially changing and unpredictable network conditions. Moreover, this dynamic optimization is entirely automatic and runs without any centralized control or any exchange of messages, using only local network feedback information. We describe the method and prove that the resulting meta-MAC protocol achieves optimal performance in a well-defined sense. Through simulation on different types of networks and with different component MAC protocols, we demonstrate that our simple and practical combination algorithm yields highly adaptive and scalable MAC solutions.