A quality of service based framework for dynamic, dependable systems
2012-05-22T15:31:17Z (GMT) by
There is currently much UK government and industry interest towards the integration of complex computer-based systems, including those in the military domain. These systems can include both mission critical and safety critical applications, and therefore require the dependable communication of data. Current modular military systems requiring such performance guarantees are mostly based on parameters and system states fixed during design time, thus allowing a predictable estimate of performance. These systems can exhibit a limited degree of reconfiguration, but this is typically within the constraints of a predefined set of configurations. The ability to reconfigure systems more dynamically, could lead to further increased flexibility and adaptability, resulting in the better use of existing assets. Current software architecture models that are capable of providing this flexibility, however, tend to lack support for dependable performance. This thesis explores the benefits for the dependability of future dynamic systems, built on a publish/subscribe model, from using Quality of Service (QoS) methods to map application level data communication requirements to available network resources. Through this, original contributions to knowledge are created, including; the proposal of a QoS framework that specifies a way of defining flexible levels of QoS characteristics and their use in the negotiation of network resources, a simulation based evaluation of the QoS framework and specifically the choice of negotiation algorithm used, and a test-bed based feasibility study. Simulation experimentation conducted comparing different methods of QoS negotiation gives a clear indication that the use of the proposed QoS framework and flexible negotiation algorithm can provide a benefit in terms of system utility, resource utilisation, and system stability. The choice of negotiation algorithm has a particularly strong impact on these system properties. The cost of these benefits comes in terms of the processing power and execution time required to reach a decision on the acceptance of a subscriber. It is suggested, given this cost, that when computational resources are limited, a simpler priority based negotiation algorithm should be used. Where system resources are more abundant, however, the flexible negotiation algorithm proposed within the QoS framework can offer further benefits. Through the implementation of the QoS framework within an existing military avionics software architecture based emulator on a test-bed, both the technical challenges that will need to be overcome and, more importantly, the potential viability for the inclusion of the QoS framework have been demonstrated.