Giada Giorgi

I received the Laurea degree in Telecommunication Engineering with honour from the University of Padova in March 2003 and the Ph.D. degree in Microelectronics and Telecommunication Engineering in March 2007.
From January 2007 to December 2007 I was awarded a one-year research contract on the analysis and measurement of performances in time constraint communication networks.
In February 2008 I was awarded a two-years research contract on the development and validation of measurement and monitoring methods for the analysis of network performances.
Since December 2008, I am working as Assistant Professor with the Instrumentation and Measurement research group at the University of Padova.
My research activity, during the last years, has been mainly focused on the aspects concerning traffic measurements and synchronization in communication and industrial networks.
The study of the statistical properties of the aggregated traffic time series, such as Self-Similarity and Long Range Dependence has allowed to understand some results discussed in the literature about the nature of these series and to go in greater depth into some significant aspects related to both the analytical models used to describe traffic and the tools used to measure their parameters.
The research activity brought contributions in the field of traffic measurements, their statistical modelling, the estimation of traffic parameters and, finally, traffic monitoring. In fact, as the rate of links increases, tools based on the tracking of protocol signalling cannot be further applied for real time monitoring, and statistical tools based on flow measurement can represent a valid alternative. It is important to note that the attention has been always focused on the identification and analysis of the physical causes of the observed behaviours in the analyzed traffic traces in order to achieve a better understanding of traffic behaviour and the factors that affect it.
The analytical method developed as the main research effort during these years, is an original approach derived from the stochastic theory of Network Calculus and based on the statistical properties of quantiles of the marginal probability distribution function related to aggregated traffic time series.
This new method, called “Rate-Interval Curves (RIC)”, was developed to overcome some limitations of the tools already available to measure statistical parameters of analytical models, where the tool based on  the Discrete Wavelet Transform is the best known.
The “RIC tool” can be applied to monitor traffic in access links of wide area networks, while for monitoring traffic in backbones, where the link capacity is very high, the computational complexity of this tool, that involves sorting operations, may become a bottleneck. o:p>
To overcome possible limitations a more computationally inexpensive tool, based on the statistical theory of Extreme Value, has been developed. This considers the asymptotic properties of the maximum, mean and minimum values series, obtained from the highest resolution aggregated time series. Since it does not depend on the particular statistical model used to describe aggregated time series, but only on general properties of these series, the tool can be adapted to work in different types of link and for different kinds of data traffic. Moreover, its computational complexity is less demanding than the RIC tool, since only the research of the maximum and minimum from a data vector is involved.
The accuracy of flow models was also taken into account, investigating a new modelling approach based on a statistical superposition of multiple flows, each having its own specific correlation structure, which results in better measurements and improved diagnostic possibilities. In particular a model based on the superposition of a main fractional Gamma process and some contaminating processes, that may affect network performances, has been studied. This model allows to approximate very well empirical traffic traces in most analysed situations and, therefore, it should prove useful during the design and optimization of systems and their components.
The most recent research activity has dealt with the development of a tool for traffic classification. The proposed method, while still referring to traffic flow, does not focus the attention on the behaviour of packets or bytes inside a single flow, but on a more efficient and faster approach. This approach is based on the behaviour of the joint probability density function between traffic statistics, calculated over non-overlapping observation windows of the same duration. This tool can be helpful in traffic monitoring to directly identify and classify network impairments.
Throughout the work, some matters concerning the instrumentation used to measure traffic in a link, have been also considered, in particular those related to the accuracy of the timestamps associated to each packet by the monitoring interface card.
Possible developments regard the study of the aspects concerning traffic measurements in hybrid networks, such as wireless networks, sensors networks and access networks for mobile communication (e.g., UMTS). In particular, flow measurements in the access link of these networks should be considered in order to provide an accurate statistical model of the traffic. This is useful both to correctly dimension network components such as access points and to develop well-suited communication protocols. o:p>
MMoreover, in high capacity link, flow analyses can help in detecting events of interest from large quantities of data, in order to promptly detect misbehaviours that may affect the quality of service.
During the last year, my interests have included the analysis of the behaviour of real-time networks and of protocols for timing and node synchronization. In industrial networks in fact, the coexistence of time-constrained cyclic data (e.g., for control) and acyclic general purpose data on the same networks, needs an accurate characterization of the behaviour of the different kind of traffic in order to analyze their impact on system performances.
A behavioural simulator of the IEEE 1588 Precision Time Protocol (PTP), based on OMNeT++, has been develop in order to assess the suitability of PTP synchronization in relation to network architecture and operating conditions, as well as for the optimization of clock servo. This simulator differs from most approaches to PTP analysis since node, link and traffic source behaviour are represented by statistical source and flow models. This allows to implement simulations where the effects of local clock instability, timestamping uncertainty and so on, can be defined individually.

Other information can be found here: my web-site