Professor and Associate Department Head, Computer Science Department, Carnegie
Analytical Performance Modeling for Power Management
in Sever Farms
Data center power consumption is increasing at an
alarming rate, having surpassed 5 billion dollars per year in costs for U.S.
data centers alone.What is most
disappointing is that much of this power is wasted.Servers are only busy 20-30% of the time on
average, but they are left on, utilizing 60% of peak power.
Capacity provisioning of server farms is non-trivial and
counter-intuitive.This talk develops
new queueing-theoretic analysis techniques which are then applied towards the
design of more efficient policies for power management in server farms.
Prof. Harchol-Balter received her Ph.D. from U.C. Berkeley
in 1996, under the direction of Manuel Blum. Her work focuses on designing new
resource allocation policies (load balancing policies, power management
policies, and scheduling policies) for server farms and distributed systems,
spanning both queuing analysis and systems implementation. She is a recipient
of the McCandless Chair, the NSF CAREER award, the NSF Postdoctoral Fellowship
in the Mathematical Sciences, as well as multiple best paper and teaching
awards, including the Herbert A. Simon Award for Teaching Excellence.
* There is no registration for this event. Seating is limited, so arriving early is recommended.