Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.446417
Title: A specification method for the scalable self-governance of complex autonomic systems
Author: Randles, Martin J.
ISNI:       0000 0001 3507 6089
Awarding Body: Liverpool John Moores University
Current Institution: Liverpool John Moores University
Date of Award: 2007
Availability of Full Text:
Access through EThOS:
Access through Institution:
Abstract:
IBM, amongst many others, have sought to endow computer systems with selfmanagement capabilities by delegating vital functions to the software itself and proposed the Autonomic Computing model. Hence inducing the so-called self-* properties including the system's ability to be self-configuring, self-optimising, self-healing and self-protecting. Initial attempts to realise such a vision have so far mostly relied on a passive adaptation whereby Design by Contract and Event-Condition-Action (ECA) type constructs are used to regulate the target systems behaviour: When a specific event makes a certain condition true then an action is triggered which executes either within the system or on its environment Whilst, such a model works well for closed systems, its effectiveness and applicability of approach diminishes as the size and complexity of the managed system increases, necessitating frequent updates to the ECA rule set to cater for new and/or unforeseen systems' behaviour. More recent research works are now adopting the parametric adaptation model, where the events, conditions and actions may be adjusted at runtime in response to the system's observed state. Such an improved control model works well up to a point, but for large scale systems of systems, with very many component interactions, the predictability and traceability of the regulation and its impact on the whole system is intractable. The selforganising systems theory, however, offers a scaleable alternative to systems control utilising emerging behaviour, observed at a global level, resulting from the low-level interactions of the distributed components. Whereby, for instance, key signals (signs) for ECA style feedback control need no longer be recognised or understood in the context of the design time system but are defined by their relevance to the runtime system. Nonetheless this model still suffers from a usually inaccessible control model with no intrinsic meaning assigned to data extraction from the systems operation. In other words, there is no grounded definition of particular observable events occurring in the system. This condition is termed the Signal Grounding Problem. This problem cannot usually be solved by analytical or algorithmic methods, as these solutions generally require precise problem formulations and a static operating domain. Rather cognitive techniques will be needed that perform effectively to evaluate and improve performance in the presence of complex, incomplete, dynamic and evolving environments. In order to develop a specification method for scalable self-governance of autonomic systems of systems, this thesis presents a number of ways to alleviate, or circumvent, the Signal Grounding Problem through the utilisation of cognitive systems and the properties of complex systems. After reviewing the specification methods available for governance models, the Situation Calculus dialect of first order logic is described with the necessary modalities for the specification of deliberative monitoring in partially observable environments with stochastic actions. This permits a specification method that allows the depiction of system guards and norms, under central control, as well as the deliberative functions required for decentralised components to present techniques around the Signal Grounding problem, engineer emergence and generally utilise the properties of large complex systems for their own self-governance. It is shown how these large-scale behaviours may be implemented and the properties assessed and utilised by an Observer System through fully functioning implementations and simulations. The work concludes with two case studies showing how the specification would be achieved in practice: An observer based meta-system for a decision support system in medicine is described, specified and implemented up to parametric adaptation and a NASA project is described with a specification given for the interactions and cooperative behaviour that leads to scale-free connectivity, which the observer system may then utilise for a previously described efficient monitoring strategy.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.446417  DOI: Not available
Keywords: QA75 Electronic computers. Computer science ; QA76 Computer software
Share: