Current network management technologies and standards are based on a centralized management paradigm. Management logic is centralized in managers who access management-information bases (MIBs) at device agents. Manageability is bounded by the number of MIB variables that may be centrally accessed and the access speed. Thus management does not scale for large, complex, high-speed or low bandwidth networks. This project aims to develop and study a novel alternative to centralized management: management by delegation. Management applications would, under this paradigm, delegate programs to agents who will apply them to monitor and control network devices. Management functions could be flexibly distributed to agents and execute in close proximity to devices. Delegation thus decentralizes management rendering it scalable for large, complex and fast networks. Success in this project would mean that the principal investigator has established novel protocols for delegation of programs and remote control of their executions; created efficient architectures for execution, coordination and security of management; established semantic models of management functions and language constructs to script them; established the performance behaviors of the proposed delegation mechanisms in comparison to current centralized management models; embedded and demonstrated these technologies within real network-management systems and tools, and exported them to academia and industry. Success would mean that the research has resulted in establishing unifying principles and mechanisms for effective distributed scalable network management.
