Speaker: Ruben Milocco (Comahue Univ., Argintina)
Title: Stochastic approach for improving energy efficiency and QoS in WSN relay selection and in Data Center management.
Date et heure : 17 avril 2019 à 16h
Abstract: The presence of both location and transmission range uncertainties require efficient relays selection in wireless sensor networks (WSN) to guarantee speed and amount of information transferred with respect to the energy used. Based on the effect that these uncertainties have on the link channel capacity, strategies to decide the next hop relay, which optimizes, in mean, the maximum rate of information transmitted with the minimum number of hops are analyzed. Topics such as, estimation of distance between nodes in a WSN; derivation and comparison of error bounds; the problem of distance uncertainties in geographic routing, and the robust geographic routing strategy by selecting nodes using metrics like loss of information and progress of information are analyzed. Data Centers (DCs) need to periodically configure their servers to meet user demands. Since the cost of the energy consumed to serve the demands is lower when DC settings are updated a priori, or proactively, there is a great interest in studying different proactive strategies based on predictions of CPU and memory requests. The amount of savings that can be achieved depends not only on the selected proactive strategy but also on user-demand statistics and the used predictors. Despite its importance, it is difficult to find studies that quantify the savings that can be obtained, due to the problem complexity. A method to compute guaranteed upper and lower saving bounds is presented. Thus, using this method together with historical records, it is possible to quantify the efficiency of different predictors as well as optimize proactive strategies. The cost reduction is evaluated with linear and nonlinear predictors that minimize optimal and robust cost functions in order to maximize the savings. Then, we apply this method to a Google dataset collected over 29 days to evaluate the benefits that can be obtained with these two predictors in this DC.
Ruben H. Milocco is professor and head of Grupo de Control Automático y Sistemas (GCAyS at the Department of the Electrical Engineering of the National University of Comahue. He is member of the Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), all from Argentina. He obtained the engineering doctorate at the National University of La Plata, Argentina and conducted postdoctoral studies at the University of Uppsala, Sweden. His research interests include filtering, estimation and identification theories with applications to communications and energy. He has more than 40 publications in high impact journals, 6 book chapters, 50 International Congresses, 60 National Congresses in the area of expertise.