Kriging based Self-Adaptive Controllers for the Cloud

You are cordially invited to attend the PhD Dissertation Defense of Alessio Gambi on Tuesday, October 16th 2012 at 09h30 in room CC-250 (Main building)

Abstract:
We propose Kriging based self-adaptive controllers to manage the allocation of resources to systems that need to provide guarantees on their quality of service at runtime while minimizing running costs.
Service providers need to adjust the configurations of their systems and the resources allocated to them to maintain acceptable level of services at runtime in face of fluctuation in the workload, dynamisms of the environment and other unexpected events.
If unsuitably configured, systems may misbehave, saturate, and violate service level agreements that lead to penalties, financial losses and damage to service providers' reputation.
Static system configurations are limited by the strong assumptions on the runtime system behavior and working conditions, and in general lead to either system over-provisioning, i.e., too expensive, or under-provisioning, i.e., too many violations.
Fixed strategies can deal well with simple systems and expected workload fluctuations, but require system experts for their setup, do not adapt, and do not scale well with system complexity.
Model based self-adaptive controllers can deal with predicted and unpredicted working conditions and can adapt.
Among the alternative models, we choose to adopt Kriging models as core elements for model based self-adaptive controllers.
Our choice is motivate by several reasons: (i) Kriging models are accurate in capturing the behavior of running systems;  (ii) they are robust against noise and inaccuracy of monitoring data, make predictions in a timely fashion, and can be retrained on-line without harnessing controllers reactiveness; (iii) they are based on a strong theory that extends traditional regression with statistical scaffoldings and that enables them to pair confidence measures with predictions; (iv) they can be used to design effective and efficient proactive controllers that account for uncertainty while planning their control actions.
We apply Kriging based controllers in the domain of Clouds at the infrastructure level, and throughout an experimental validation, we show that Kriging models are accurate, fast and flexible enough to be suitably employed by model based self-adaptive controllers.  We compare Kriging based self-adaptive controllers against state-of-the-art solutions and we conclude that our controllers are efficient, effective and more generally applicable than the other considered solutions.

Dissertation Committee:

• Prof. Mauro Pezze', Universita' della Svizzera italiana, Switzerland (Research Advisor)
• Prof. Antonio Carzaniga, Universita' della Svizzera italiana, Switzerland (Internal Member)
• Prof. Cesare Pautasso, Universita' della Svizzera italiana, Switzerland (Internal Member)
• Prof. Marin Litoiu, York University, Canada (External Member)
• Prof. Nenad Medvidovic, University of Southern California, USA (External Member)