Opportunities for
Control Engineering
In Computing Systems
Dr.
Joseph L. Hellerstein
Modern
society has a profound reliance on computing systems. These are complex structures of multiple (often
distributed) components with non-linearities and stochastics. While queueing
theory is widely used to address
steady-state stochastics, practitioners rarely address dynamics (e.g., limit
cycles, long settling times), often with unfortunate results. In particular,
control engineering is almost never employed in practice. Over the last four
years, my colleagues and I at IBM in collaboration with others at universities
(e.g., Dawn Tilbury at University of Michigan) have demonstrated the practical
value of applying the basics of control engineering (e.g., linear,
deterministic, time invariant systems) to real world computing systems. Some of
this work has been incorporated into IBM products. This talk will provide our
perspective on the opportunities for control engineering in computing systems.
This talk will also be presented as a tutorial at ACC 2004.
3:30 Ð 4:30 p.m.