Abstract

This presentation will discuss the application of autonomic computing
to load-balancing for a multi-tiered auction web site.  Autonomic
computing systems are able to adapt to changing environments (such as
changes in the workload intensity or component failures) in a way that
preserves high-level operational goals, such as service level
objectives.  The autonomic load balancer of the web site divides the
bottleneck server tier into clusters, each of which is dedicated to a
certain priority class of users.  The autonomic load balancer
dynamically adjusts resource allocations to the clusters in an effort
to maximize a utility function based on response time and bid
throughput.  To reduce switching costs, the load balancer also
considers load balancing policies that redirect a percentage of
requests intended for one cluster to a different cluster.  To make
allocation and policy decisions, the autonomic load balancer uses an
efficient heuristic search to explore a utility landsc ape generated
by the predictions of an analytic performance model.  Another key
contribution presented here is a novel method for the random
generation of realistic stress tests.  Using this stress test method,
the autonomic load balancer is assessed against both a round-robin
load balancing approach and a dedicated cluster approach.

Speaker Bio

John Ewing is currently a PhD student in the Computer Science
Department at George Mason University.  From 2001 to 2005, John worked
at the Defense Information Systems Agency conducting capacity planning
studies and performance troubleshooting of large, distributed software
systems with millions of users.  From 1999 to 2001, John worked as a
government IT contractor analyzing performance of computer systems and
developing prototype software.  John received his Masters in computer
science from the Illinois Institute of Technology in 2003 and his
Bachelors in chemistry from the University of Richmond in 1997.