Rigorous Systems Research Group (RSRG) Seminar
Annenberg 213
Dynamical Models for P2P and Energy Regulation Systems Inspired by Queueing Theory
Andres Ferragut,
School of Engineering,
Universidad ORT Uruguay,
In this talk, I will present some fluid models for P2P file sharing systems and power regulation problems. The main common point in these systems is that a series of jobs must be served by a varying capacity, which may be determined by current occupation (P2P) or by an external signal (power regulation). Job sizes may have a general distribution, and they may have service deadlines that the system must comply with. The focus of the models is in describing the macroscopic behavior of the systems involved, ignoring the finer microscopic details of the service policies in use.
The large scale of the systems involved suggests the use of fluid models, representing the average behavior of the underlying queueing system. For the P2P case, a PDE model is derived that incorporates the fact that peers are interested in a common file, providing results on equilibrium stability, variability and transient behavior.
For the regulation problem, a similar model is presented inspired by queueing theory, and some of the consequences on power regulation are analyzed. In particular, we present a macroscopic design rule for exploting load deferrability to offer ancilliary regulation services. Connections with interesting queueing problems will also be discussed.
The large scale of the systems involved suggests the use of fluid models, representing the average behavior of the underlying queueing system. For the P2P case, a PDE model is derived that incorporates the fact that peers are interested in a common file, providing results on equilibrium stability, variability and transient behavior.
For the regulation problem, a similar model is presented inspired by queueing theory, and some of the consequences on power regulation are analyzed. In particular, we present a macroscopic design rule for exploting load deferrability to offer ancilliary regulation services. Connections with interesting queueing problems will also be discussed.
For more information, please contact Sydney Garstang by email at [email protected].