Autonomic Control Frameworks
Dr. Sherif Abdelwahed
In this research project, we will develop an autonomic performance and security management technology that integrates system control, optimization and security analysis tasks into a common model-based framework that enables distributed computing systems to adapt efficiently to variations in load requirements and identify potential security intrusions and defend the system to maintain functionality. We will develop a computationally efficient model for distributed computing systems which provides effective estimates of system behavior through aggregation of workload and resource utilization data as well as the stochastic probability of network intrusion for each critical interface in the system. The proposed integrated security management and adaptive performance control framework will unify the two tasks of managing systems performance in normal operating conditions as well as under security attacks.
The expected outcome of this research initiative is a set of technologies to convert a significant number of system management tasks into systematic and semi-automated processes using concrete mathematical models, and proven optimization techniques based on control theory. The proposed research has the potential to significantly reduce the operating cost of current and future large-scale distributed systems as well as improving the resilience, reliability and Quality of Service of the applications hosted on these systems. We will validate the solutions developed in this project and demonstrate their effectiveness when applied to selected problems in distributed computing systems management. We will demonstrate a fully implemented control framework for real-time management of a multitier architecture comprising several hundred servers using the High Performance Computing Collaboratory facilities at Mississippi State University.