Interdisciplinary Research: Roles for Self-Organization

• Authors: Mark Jelasity;Ozalp Babaoglu;Robert Laddaga;Radhika Nagpal;Franco Zambonelli;Emin Gun Sirer;Hakima Chaouchi;Mikhaill Smirnov
• Venue: IEEE Intelligent Systems
• Year: 2006

Interdisciplinary Research: Roles for Self-OrganizationMárk Jelasity and Ozalp Babaoglu, Università di Bologna; Robert Laddaga, Massachusetts Institute of TechnologyNo generally accepted principles and guidelines currently exist to help engineers design local interaction mechanisms that result in a desired global behavior. However, several communities have developed ways of approaching this problem in the context of niched application areas. We've invited representatives of several communities to review the role of self-organization in their work.Self-Organizing Shape and Pattern: From Cells to RobotsRadhika Nagpal, Harvard UniversityMulticellular organisms function as a whole even though they're essentially a colony of cells that are constantly dying and being replaced. If we wanted to create such a systems, what would we tell the cells to do? The Massachusetts Institute of Technology's Amorphous Computing project conducts research to answer this question.Self-Management and the Many Facets of "Nonself"Franco Zambonelli, Università di Modena e Regio EmiliaNearly all "self-*" features of self-managing systems--self-configuration, self-adaptation, self-healing, and so on--consider human beings as "nonself" in relation to the system. Autonomic computing inherits this architecture and merely replaces human managers with digital components. Biologically inspired self-organizing systems are limited in their current development. An ecological perspective on the self-managing problem injects "manager" components into a system with knowledge and capabilities beyond those available to normal components.Heuristics Considered Harmful: Mathematical Optimization for Resource Management in Distributed SystemsEmin Gün Sirer, Cornell UniversityDistributed systems often pose difficult resource management problems that involve partitioning a critical resource--such as bandwidth, storage, or computational elements--between competing tasks. Traditionally, such problems are resolved using custom, domain-specific heuristics. Yet heuristics are not robust to fluctuations in load characteristics, nor do they enable the system designer to reason definitively about emergent system properties after deployment. Mathematical optimization offers a more principled approach to distributed systems resource management, and it's ideally suited to resource allocation problems. The author outlines a general approach based on analytical modeling, optimization, and implementation practicalities. Its application to several domains has yield qualitative improvements in performance and robustness.Autonomic Communication: Business-Driven RevolutionHakima Chaouchi, National Institute of Telecommunication and Mikhaill Smirnov, Fraunhofer FOKUSAutonomic communication is an emerging concept to handle the ever-increasing complexity of network control and management. The authors describe the concept in the context of known network self-management features like automation, adaptation, and reconfiguration. They argue that autonomic communication is a natural evolution of network self-management but enables a revolution in communication services. They provide a high-level roadmap toward a totally autonomic communication system that is governed by user and business processes.