Most relevant research achievements
Construction of optical buffers
The design of a first in first out (FIFO) optical buffer capable of emulating a shared queue is investigated. The construction, based on optical switched fiber delay lines (SDL), consists in a space switch in which SDL lines connect its outputs back to its inputs, thus acting as a shared buffer. It is proven that such structure exactly emulates the behavior of 1x2 shared FIFO queue.
Coverage Optimization of 3G Systems
Mathematical programming models are proposed for supporting the decisions on where to install new base stations and how to select their configuration (antenna height and tilt, sector orientations, maximum emission power, pilot signal, etc.).
Estimation of traffic long-range dependence
For first time, the Modified Allan Variance (MAVAR) and a new Modified Hadamard Variance (MHVAR) are proposed to estimate the Hurst parameter H of LRD traffic series or the exponent a of data with 1/f^a power-law spectrum. MHVAR generalizes the principle of MAVAR to higher-order differences of input data. Extensive simulations show that MAVAR and MHVAR achieve significantly better accuracy and confidence in H estimation than popular wavelet-based methods.
Mobile network traffic analysis
Time-correlation of answered call arrivals in sets of real GSM telephone traffic data is investigated, for the first time by taking advantage of the excellent capability of the Modified Allan Variance in discriminating power-law noise types. The empirical statistics of call arrivals proves excellent consistency with the ideal Poisson model with diurnal variable rate, also by chi-square test evaluation.
Multimedia applications transport analysis
A general closed-form analysis of end-to-end network delay has been proposed. The effect of traffic control mechanisms such as bucket-based policers has been studied in a very general framework, proving that the effect of a classical policing device on traffic is intrinsically limited and cannot eliminate long-range dependence in every case. A spin-off of the theoretical framework set up in the described work, has resulted in a new general approximated analysis of queuing delay in GI/G/1 queues, offering better precision than known approximations.
Optimal Overlay Networks
Mathematical programming models are proposed for the user assignment problem, the traffic routing optimization and the dimensioning of the capacity reserved on overlay links in SONs. Numerical results show that the proposed approach is able to solve the problem to the optimum even for large-scale networks.
Radio Planning of WLANs
Mathematical models are proposed to tackle the WLAN planning problem. The single-channel and the multiple-channel WLAN planning problems are considered. To evaluate the quality of the proposed models, the optimal solutions for synthetic network instances are obtained, and heuristics are proposed to get suboptimal solutions in a reasonable computing time. The networks planned according to the proposed approach feature higher efficiency than the ones planned using classical models.
SLA-Aware provisioning in optical core networks
Automatized management in optical networks allows customers to specify on-demand these QOS requirements in a Service Level Agreement (SLA) to be guaranteed by the service provider. Thus, service-level specifications are made available to the control plane to perform effective provisioning. We explored various cross-layer design opportunities that arise when the SLA terms are specified for on-demand connection requests, considering, e.g., availability, duration or setup-delay tolerance of the connections.
Topology Optimization of Wireless Mesh Networks
Novel optimization models are proposed for planning Wireless Mesh Networks (WMNs). The mixed integer linear programming models allow to select the number and positions of mesh routers and access points, while accurately taking into account traffic routing, interference, rate adaptation, and channel assignment. The optimal solutions of three problem formulations are proposed and a relaxation-based heuristic is introduced for large-sized network instances.
WDM network design by ILP models based on flow aggregation
Integer Linear Programming (ILP) is the most used exact method to plan and optimize WDM networks. Unfortunately, even for small networks, solving ILP may require too much computational effort. Our formulation, aimed at providing computational efficiency, exploits flow aggregation and consists in a new ILP formulation; this allows us to reach optimal solutions with less computational effort compared to other ILP approaches.