Professor Emeritus and Distinguished Research Professor
My long-time interest is in efficient modeling of large-scale distributed computer systems, using layered queueing to represent resource contention. This can be broken down into subtopics
(1) a fast approximate solution tool called LQNS (work continued by Prof Greg Franks),
(2) methods for finding the structure of models from request traces, and for finding their parameters by nonlinear estimation and regression techniques,
(3) deriving performance models from software designs in UML (the PUMA system and the MARTE profile for UML annotation, with Prof Petriu), and
(4) model-based optimization for better design, configuration and deployment, most recently for cloud management (with Prof Chinneck and Litoiu).
Recent research is focussed on the last topic, which can be summarized as “exploiting models for design and management”.
An information page about layered queueing is here and our LQNS solver for layered queueing systems is here . An information page about regression techniques for fitting LQ models is here .
Positions for new graduate students:
I am looking for students to study cloud optimization and models of software.
List of Publications
RADS Lab Home Page
UML Performance Profile: Documents