Self-adaptability, resilience and vulnerability in autonomic communications with biology-inspired strategies

Publication Type:
Issue Date:
Full metadata record
Files in This Item:
Filename Description Size
01Front.pdf2.72 MB
Adobe PDF
Thumbnail02Whole.pdf114.8 MB
Adobe PDF
NO FULL TEXT AVAILABLE. This thesis contains 3rd party copyright material. ----- Autonomic communication networks, as a high-level goal-oriented self-managed paradigm, are emerging as a promising solution to complex adaptive networks in communication fields. It aims at releasing human network operators from low-level tasks. Over the years, researchers have proposed a number of solutions in order to achieve such a goal-driven self-managed paradigm, such as ontology structuring and engineering of the knowledge base to help with decision-making. However, these methodologies are struggling to cope with large computational complexities, incomplete (or inaccurate) and unsynchronised knowledge in distributed network environments. In comparison, biological systems have demonstrated their instinctive capability to learn and adapt. Over years' evolution and natural selection, social insects and biological organisms have developed relatively easy and efficient mechanisms to thrive in hostile, dynamic and uncertain environments. The basic philosophy is that local agents without knowledge of system goals and objectives, make decisions based on local conditions. But system-level resilience and robustness are presented as emergent properties. A lack of literature exists in achieving the autonomy via distributed biology-inspired learning and adaptation strategies. This thesis aims at contributing to the state of the art. Biological metaphors are applied to the design and implementation of bio-inspired learning and adaptation systems. The necessary architecture, multi-agent framework, information model, algorithms for enabling learning, adaptation strategies and systematic models as well as corresponding protocols are described. In addition to the introduction (Chapter 1) and literature background reviews (Chapter 2), the first part of the thesis (Chapters 3 and 4) accumulates the theoretical foundation for the rest of the thesis. This foundation consists of a generic architecture for complex network management, a unified multi-agent framework inspired by the biological behaviour, as well as the O:MIB model, (served as an information model to aggregate distributed network information). It is designed and implemented by the hybrid O:XML language with a customised Java compiler. The second part of the thesis (Chapter 5) proposes a new biologically inspired threat awareness strategy, inspired by immunology, to reconfigure file access right. An integrated system is constructed where a distributed threat awareness system integrates with a constrained ant-based self-adaptive system. The experimental results validate this methodology, and suggest a statistical bound for operators to set as a vulnerability level in practice. With an improved self-adaptive ant metaheuristic, the third part (Chapter 6) present's a case study. The constrained ant-colony adaptation algorithm is applied, as a decentralised adaptation strategy, to adapt the overall integrated system towards the self-configuration of network services based on an SLA and system objectives. Its convergence and robustness performance are compared with the centralised adaptation benchmark PBIL. This thesis concludes with an exploration of vulnerability, autonomy and adaptation capability. This enables a clear and in-depth understanding of their relationship, and hence high-level manipulations of them assist in autonomic communications. The biology-inspired self-learning and adaptation algorithm and the threat awareness model work together for an integrated system. The good performance is reported in terms of convergence and robustness for the service-configuration process of autonomic communication networks. This integrated system can be extended to other distributed IP-based networks.
Please use this identifier to cite or link to this item: