Design of an adaptive computing architecture for managing interactions in heterogeneous defence networks
Bruce Davis McClure · The University of Queensland · 2003
Today, to fulfil the operational requirements of modem battlefield interoperability, defence organisations require the deployment of distributed applications such as video on the Web, situational awareness and collaborative planning on heterogeneous networks. Defence networks comprise communication technologies ranging from Giga-bit fiber-optic links to satellite channels and HF radio links and may be subject to physical or electronic attacks on communications. Existing system and application architectures, which typically assume LAN quality networking, do not work well for heterogeneous defence networks or perform well as available network QoS degrades. There is, therefore, a real need to be able to manage the use of network and communications resources so that it is possible to dynamically adapt to changes in service requirements and network conditions. There has been extensive research in the area of Quality of Service (QoS) management and adaptive computing for mobile applications. Approaches to the problem include caching, filters, proxies, modified RPCs and application adaptation. None of this work, however, addresses the full scope of the problem. The defence environment requires a comprehensive architecture for resource management that is capable of selectively providing guaranteed service according to an over-arching policy. A piece-meal solution that relies on the applications and services working cooperatively is not sufficient. Policy based resource management is a function that needs to be an integral part of the distributed computing infrastructure. This thesis proposes an Adaptive Computing Architecture to manage enterprise-wide network resources in heterogeneous defence networks according to predefined adaptive policies in an ODP environment. In this thesis, an ODP-based architectural framework for adaptive resource management in heterogeneous defence networks which includes a model for enterprise-wide adaptability policies. The model provides a more flexible approach than other policy based approaches for the Internet network. The adaptability policies are represented in a policy language whereas other approaches provide only a list of policy options to select from or require re-compilation of the system to modify policies. The thesis defines a model of network topology and QoS for representing networks at various levels of abstraction to support adaptive decisions. This model is flexible enough to represent networks with significant detail where required and able to abstract over large portions of networks that do not need to be involved in adaptive decision-making. Several QoS models for representing application QoS requirements and agreed QoS contracts are set out in the thesis. These models have been refined to provide sufficient power to describe requirements, provisions and contracts anticipated for open, distributed defence applications and sufficiently specific to facilitate unambiguous interoperability with other applications and architectures. The thesis demonstrated, via qualitative assessment, that the proposed architecture improved QoS delivered to critical applications through the application of adaptability policies to manage network resources in heterogeneous defence networks.