Reinforcement learning in policy-driven autonomic management

In order to effectively manage todays complex systems, system administrators are turning to automated solutions. Policy-driven management offers significant benefits since the use of policies can make it more straight forward to define and modify systems behavior at run-time, through policy manipulation, rather than through re-engineering. The use of policies within autonomic computing allows system administrators to embed existing knowledge into policies and thereby drive autonomic management. Equally important, however, is a need for autonomic systems to adapt the use of these policies to deal with not only the inherent human error, but also the changes in the configuration of the managed environment and the unpredictability in workloads. This paper reports on the use of reinforcement learning methodologies to determine how to best use a set of enabled policies to meet different performance objectives. The work is presented in the context of an adaptive policy-driven autonomic management system.