Modeling and Engineering Self-Organization in Complex Software Systems

Describing, understanding, and modeling the emergent behavior of self-organizing software systems remains an open challenge. Such systems can solve problems in computing domains where traditional, centralized models are impractical or problematic, including ubiquitous and pervasive computing, peer-to-peer networks, large-scale grids, and Ultra-Large-Scale Systems. Self-organizing approaches have demonstrated great promise in building adaptive behavior into decentralized systems, enabling cooperative, autonomous self-management and the exploitation of the heterogeneity of system components. My investigation of self-organizing software systems has revolved around Myconet, an unstructured overlay protocol for peer-to-peer networks. Myconet takes inspiration from fungal growth patterns in order to build an efficient self-optimizing superpeer topology that can also rapidly self-heal in response to damage orattacks. Myconet has proven to be flexible, and has been used as a platform for the development of other self-organizing applications in large-scale distributed systems, including load-balancing in distributed service networks (Mycoload), and detection and mitigation of attacks against the overlay (Hormone-Inspired Topology Adaptation Protection [HITAP] and Self-Organized Degree Adaptation Protection [SODAP]). Each extension has given additional insights into the self-organizing dynamics of such systems, but has also shown the limitations of ad hoc approaches to the design and analysis of new applications. These experiences have led me to investigate formal tools and models that may provide the designer of a self-organizing system with early and accurate insight through augmented analytical power. This research selects a small set of synergistic modeling techniques, and builds an integrated approach to modeling for the design and validation of self-organizing software systems. These tools are used to model the core Myconet platform and its currently developed extensions, particularly focusing on the SODAP layer which provides self-protection features to a superpeer-based P2P overlay network. Once established, this modeling approach can be applied to the principled design of further Myconet extensions, as well as other self-organizing systems, thus advancing the understanding of how to model and engineer self-organization in software systems.