On Observation of and Interaction in Open Distributed Systems

Sammanfattning: The thesis presents the results of research in computer science, as applied within a series of industrial projects. All of these projects have been part of an iterative and adaptive development process, from which we can draw conclusions in three different aspects. The first aspect concerns the processes of the projects as carriers of scientific ideas from the university to an industrial environment. The second aspect concerns findings related to principles of observation of and interaction in open distributed systems, and some phenomena that have been shown through explorative experiments. The third aspect concerns the relevance of these results to the selected application area; systems for network enabled capabilities. The series of demonstrators mentioned above are based on the model of open computational systems (OCS). It has been shown that the OCS model is applicable for the engineering and development of systems for network enabled capabilities. However, we have also seen that such systems sometimes show indeterministic behaviours, possibly leading to safety problems. Therefore the research objectives of this work have been related to increasing the understanding of the principles of observation and interaction in open distributed systems. Our results lead to the following two main conclusions of this thesis: The first conclusion is that there exist inherent limitations as of to the degree of precision with which measurements can be made in distributed systems, such as command and control systems. These limitations are related to quantum properties of the cyber world. This uncertainty phenomenon also makes it impossible to precisely assess value metrics associated with high level conceptual models of command and control. The second conclusion is that distributed systems also exhibit relativistic properties, caused by the limited transfer speed of data in transmission networks. As a consequence of this we suggest that the age of a message at arrival is a better descriptor of network interaction characteristics, than the time a message is sent. We also introduce the coherence interval, which we suggest is used as a basis for a metric of information reach in networks. The coherence interval provides indications of how often messages are to be sent from a node, in order to maintain an acceptable age of messages at arrival. If interactions between nodes are occurring slower or faster than suggested by the coherence interval, the resulting observations as arriving at other nodes become either irrelevant or meaningless.