|
||||
|
Networks
Evolutionary autonomic design framework for self-organizing Future Internet | Permalink The Future Internet is expected to comprise a plethora of self-organizing autonomic devices1 of varying capabilities and levels of intelligence, collaborating to accomplish actions that cannot be achieved otherwise. As the number and complexity of interactions increases, an inherent tradeoff between the gains and costs of collaboration emerges. In a dynamic and operationally efficient way, systems must strike an appropriate balance between two basic challenges. The first is that given behavioural (e.g. application-oriented) characteristics and requirements, what structure, subject to constraints, optimizes performance? The second is that under the constraints of a given structure, how well can a desired behaviour be executed? Following this paradigm, we anticipate that the Future Internet-connected world will be not simply a collection of linked asynchronous nodes but a multi-loop feedback network of distributed asynchronous dynamic systems, active databases and knowledge bases. Network science has lately grown into a multi-discipline research area devoted to the identification, modelling and analysis of complex networks and interactions in all manner of natural and artificial settings.2 Historical advances and evolution in contemporary network science have led to the identification and practical employment of a three-level model: physical, logical and social. At the ground level are physical networks, where node associations correspond one-to-one to physical connectivity. At the second level are logical networks, based on virtual associations and connectivity among peers. The third level, denoted social networks, involves more complex interactions and takes into account mainly unpredictable or hidden associations, including application-specific requirements. Traditional network design flow has not yet achieved the full potential of this three-layer model. For example, although logical networks are built on top of physically interconnected nodes, their connectivity properties may differ significantly, as may be seen in peer-to-peer or wireless ad-hoc networks. In fact, network designers often redesign the physical infrastructure of networks to meet requirements posed at the logical level, especially in wireless systems, where reconfiguration is technically and economically viable. Similarly, logical networks impact social networks, as can be readily observed from studying preferential attachment in the derivation of social networks from logical ones. Several features of social networks are appropriate when addressing some of the deficiencies of the physical topologies now in use (e.g. wireless multi-hop networks). Of particular interest are the preferential attachment rule, the small-world phenomenon, and scale-free structures,3 which tend to manifest themselves via such indicators as a small average path length, high clustering and connectivity robustness in the presence of random failures. Several efforts, including ours, have suggested mapping these social features onto communication networks.4, 5 It has, for example, been shown that small-world features can optimally address such problems as distributed consensus in wireless multi-hop networks.6 A key part of our research has been to devise a distributed autonomic framework that efficiently achieves these objectives without significantly impacting other critical characteristics of the network infrastructure. An example is shown in Figure 1, which was obtained by an inverse topology control mechanism.5 Figure 1. Model of network evolution. (a) Initial network. (b) Links added. (c) Nodes added. (d) Finally induced network. 
Achieving this goal in a dynamic wireless multi-hop network requires that researchers:5, 7 focus on closing the loop between social and physical networking in the design paradigm; exploit ways in which social knowledge and the features of online social networks can improve the characteristics of physical communication networks; demonstrate feasibility by infusing the desired properties of online social structures (small-world effect, power-law-like degree distribution) into existing dynamic infrastructures, such as the core structure of wireless multi-hop networks; use inverse topology control-based techniques to modify communication links in multi-hop networks; analyse interactivity through a continuum-theory-based framework,8 hyperbolic graph theory,9 game theory10 and systems of ordinary differential equations;5 and identify the underlying research challenges that must be addressed for a more holistic treatment and exploitation of the vision of evolutionary design. Thus, in order to set the norm for future holistic network design approaches, what is required is not simply an extension of the traditional design hierarchy but rather the development of a novel framework. This not only takes into account both directions of the interactions among physical, logical and social network viewpoints, but more importantly it aspires to close the loop along the traditional design and analysis hierarchy, potentially leading to an autonomic evolutionary loop.1 According to this paradigm, the network's progressive advance to more complex features will spur changes in the fundamental operations from which those features emanate, igniting a new evolutionary loop that produces additional advances, as seen in Figure 2. Figure 2. A proposed autonomic design framework for context-aware and socially driven heterogeneous Internet-connected object networks. As the two loops indicate, evolution proceeds in both directions. 
In order for an underlying network architecture to improve its own efficiency evolutionarily, by dynamically modifying its structure and features, it must have increased capabilities, including awareness of its environment, state and type. The whole universe of interactions within the digital and physical worlds must be traversed by following a bidirectional path: bottom-up, for aggregating and processing data collected by physical sensing objects and using it for situation awareness and decisionmaking at higher levels; and top-down, for compiling application demands into performance-improving topology modifications (sensing requests and control actions) over the logical and physical infrastructure, taking into account the properties and features of social networks. and technical editor for IEEE Wireless Communications Magazine. References
Stay Informed
|
||||
|
||||