Emergence in active networks
thesisposted on 29.05.2014 by M. Shirantha de Silva
In order to distinguish essays and pre-prints from academic theses, we have a separate category. These are often much longer text based documents than a paper.
Any complex system may potentially exhibit unpredicted and undesirable behaviour as a result of certain combinations of input stimuli. An Active Network, being a communication network in which user requested operations are undertaken in the netwOIk nodes themselves, is a candidate to exhibit such behaviour. For example, resource utilisation will be influenced by the specific combination of activities triggered by the users and may develop undesirable characteristics such as a self-sustaining profile. Conventional simulation tools do not detect such characteristics. This thesis proposes a solution based on a Petri-Net model in which the resource utilisation of the Active Network is abstracted above the link level communication element. It is then suggested that a certain type of Emergence in resource utilisation may manifest itself as Self-Similarity. The Hurst Parameter (H) of the resource utilisation profile for each node in the network can then be used to identify the presence of this characteristic. The RlS Statistic is used to estimate sets of H values for a range of different Active Application scenarios. It is subsequently seen that a self-sustaining resource utilisation profile (termed a "Cascading Effect") occurs when a significant subset of the nodes display high values of H. This thesis takes the view that Emergence in Active Networks is a problem that has to be approached with a global comprehension of the system as opposed to the conventional approach of a piecemeal development of solutions. This view is reinforced by the hypothesis that an Active Network is a Complex System and Emergence is noncomplex self-organisation within it. It proposes that the high-level abstraction of the Active Network forms a view by which global comprehension can be obtained and is used for the detection of anomalous behaviour (Le. Emergence). The key enabler for self-organisation is proposed to be 'the resources' within the Active Network nodes and hence the detection technique was focused on the utilisation characteristics of these.
- Mechanical, Electrical and Manufacturing Engineering