Achieving Self-management in a Distributed System of Autonomic BUT Social Entities

This paper presents a novel self-management model for resource allocation in an autonomic system (AS) comprised of individual, but social, autonomic entities (AEs). Each AE is associated with an interdependent utility function that, not only models its utility over its resource allocations, but also depends on other AEs allocations and, hence, the global AS welfare. Pervious utility-based approaches are limited to representing the AS as a set of independent AEs that aim at self-optimizing their performance unaware of other AEs' behavior. In contrast to these dominant approaches, the proposed scheme efficiently models various social behaviors, such as cooperation, selfishness and competition, among those AEs to dynamically change the overall resource allocations in different scenarios such as in the case of anomalies or varying service demands. These behavior patterns are incorporated into the utility function of each AE which is composed of two components, local and global utilities. The former reflects the AE's utility of its resource consumption while the latter is dependent on the other AEs' consumptions. By controlling these utilities, AEs create a social community where they lend/borrow resources and reward/punish other well/mal- behaving AEs. Experimental results demonstrate that creating such a social AS is more efficient than simplified systems of independent utilities.