Self-Organization in Peer-to-Peer Systems

Peer-to-Peer systems have often been advertised to as being “self-organizing”. This has become a somewhat elusive property with meaning and significance being far from clear. Answers sought in this research therefore focused on questions like: How well is self-organization understood in terms of peer-to-peer systems? How much self-organization is there really? How can, if at all, mechanisms of self-organization be exploited for control and management of peer-to-peer systems? Self-organization is closely related to the phenomenon of structures that emerge out of activities by groups of lower components. The interactions among those components, thereby, are governed by rules purely based on local information. There is no external ordering influence that could impose the emergent properties that appear at the global level. Emerging structures are driven by positive feedback information that amplifies fluctuations. To balance systems, on the other hand, negative feedback information plays a crucial role for selfregulation. Examples of negative feedback would be depletion of resources due to saturation or competition. As a result, structure emerges as a function of an interaction between positive and negative feedback information. The emerging structure often exhibits so-called smallworld characteristics where a few components have relations to a very high number of other components and many components have only very few relations to a few other components. Self-organizing systems operate in a critical state that allows them to respond adaptively, but not chaotically, to some stimuli. Typically, there exist several such semi-stable, critical states such that bifurcation, that is, state change, may occur in response to parameter variations. Multi-stability in critical states is crucial for adaptive, self-organizing systems. Selforganization arises in systems with multiple iterative interactions among components that behave according to some rules-of-thumb. This results in a recursive process with cascades of interacting mechanisms where complexity unfolds progressively. Self-organization has to be distinguished from systems that are centralized and from systems for which there are blueprints, recipes or templates available. Here components are “unaware” of any global situation and complex group behaviour evolves from a component-level behaviour that is much less complex. The emerging structure is non-predictable and nondeterministic. Different group-level patterns could thus arise from identical conditions. Diversity of responses could result in qualitative and quantitative changes in groups while components do not change behaviour. This is part of stimulus-response patterns in critical state systems. The environment, thereby, acts as strong constraint that shapes the selforganization process. Scalability, flexibility, evolveability and self-management without any central control are among the most prominent and attractive properties of self-organizing systems. This is particularly attractive from the perspective of the Internet, where limitations in scalability and flexibility have been increasingly manifesting themselves as major problems and challenges. This is in particular so due to prevalent client-server architectures with their implied limitations on the server-related resources on the one hand and plenty of unused resources at the client side on the other hand. The recent emergence of peer-to-peer systems has countered that trend. Here interacting end systems (peers) jointly use resources in end systems in a more or less decentralized fashion. Dagstuhl Seminar Proceedings 04411 Service Management and Self-Organization in IP-based Networks http://drops.dagstuhl.de/opus/volltexte/2005/86 Copyright and Reference Information: This material (preprint, accepted manuscript, or other author-distributable version) is provided to ensure timely dissemination of scholarly work. Copyright and all rights therein are retained by the author(s) and/or other copyright holders. All persons copying this work are expected to adhere to the terms and constraints invoked by these copyrights. This work is for personal use only and may not be redistributed without the explicit permission of the copyright holder. The definite version of this work is published as [·] Hermann De Meer. Self-organization in peer-to-peer systems. In Matthias Bossardt, Georg Carle, David Hutchison, Hermann De Meer and Bernhard Plattner (editors), Service Management and Self-Organization in IP-based Networks, Volume 04411 of Dagstuhl Seminar Proceedings, Dagstuhl, Germany, 2005. Schloss Dagstuhl, Leibniz-Zentrum fuer Informatik GmbH. See http://www.net.fim.uni-passau.de/papers/DeMeer2005e for full reference details (BibTeX, XML).