Quand: le jeudi 9 novembre 2017 à partir de 11h
Où: A Télécom SudParis, en amphi. 11
Breaking the dimensionality curse
in multi-server queues
Alexandre Brandwajn,
Basking School of Engineering, University of California at Santa Cruz, USA
and
PALLAS International Corporation, San José, CA, USA
Ph/Ph/c and and Ph/Ph/c/N queues can be viewed as a common model of multi-server facilities. This talk presents a recently developed simple approximate solution for the equilibrium probabilities in such queues. The proposed solution is based on a reduced state description in order to circumvent the well-known and dreaded combinatorial growth of the number of states inherent in the classical state description. The number of equations to solve in our approach increases linearly with the number of servers and phases in the service time distribution. A simple fixed-point iteration is used to solve these equations. Our approach applies both to open models with unrestricted buffer size and to queues with finite-size buffers and features generally good accuracy. This work was done jointly with Thomas Begin of ENS Lyon.
Alexandre Brandwajn holds a Ingenieur Civil des Telécommunications degree from the Ecole Nationale Superieure des Télécommunications in Paris, and a Docteur d’Etat in Computer Science degree from the University of Paris VI. He worked as researcher at the Institut de Recherche en Informatique et Automatique (IRIA), France, then he was on the faculty of the Ecole Nationale Superieure des Télécommunications in Paris where he directed a project in adaptive computer architecture. Later he joined Amdahl Corporation in Sunnyvale, California, where he was a Senior Computer Architect, and then Manager of Systems Analysis group. Since 1985 he is a professor of Computer Engineering at the University of California at Santa Cruz. His current research interests include efficient solution of systems with large state space, application of conditional probability in the solution of performance models, models of virtualized systems, as well as efficient solution of priority systems.