SOC Research Area:
Simulation and Optimisation of Service Systems

To simulate complex systems of services so as they may be both analysed and optimised.

 

Background

As service-oriented systems become complex the number of possible ways in which interactions can be organised between the services becomes infeasible for any form of exhaustive evaluation. Simple techniques that independently build sequences of interactions, each of which is optimised in isolation, are fast but rarely give optimal results. In fact the overall result obtained is often poor as there is no capacity to make a sub-optimal choice at one step because this will enable a later interaction to be made that more than recovers the lost ground.

Service-oriented systems rarely remain unaltered for any lengthy period of time. Rather the objectives, requirements and constraints on the system are continually altering, mostly in a small way but with occasional large changes. Any technique that is to optimise a complex service system must be able to handle these conditions.

While many techniques exist to simulate and optimise simple systems with requirements do not vary significantly with time and typically do a very good job. The techniques being developed at Swinburne, while they can efficiently handle to simple situations, are also suited to complex service systems that are required to continually adapt to ever changing requirements. The key design criteria for these techniques are that they are both efficient and also scale controllably with problem complexity.

 

Research Issues/Capabilities

The research into Simulation and Optimisation of Service Systems addresses several key challenges of Service-Oriented Computing, including:

	Modelling service systems that are too complex to permit formal modelling. This can be done from no more than a list of examples of system performance.
	Optimising model parameters for complex systems of services so as to optimise their performance (as specified by a user defined set of key performance criteria).
	Detecting anomalous behaviour in the performance of such complex service systems such as conditions that produce inappropriate performance.

 

 

The research is conducted by a team of internationally renowned researchers and postgraduate research students contributing expertise in a range of technologies including intelligent systems, complex systems, intelligent agents, workflow technologies and web/grid service engineering, from five closely related research programs:

	Complex Intelligent Systems led by Prof. Tim Hendtlass
	Intelligent Agents and Multi-Agent Systems led by Prof. Ryszard Kowalczyk
	Workflow Technology led by Prof. Yun Yang
	Component Software and Enterprise Systems led by Prof. Jun Han
	Information Systems led by A/Prof. Judy McKay

 

Selected R&D projects

A number of industrial projects have been successfully completed with both industry and government over the past five years. These have ranged from improved product flow within a complex industrial plant to the detection of anomalous events in high volume data transactions. Both contact and collaborative research projects have been undertaken, but confidentiality agreements do not allow further information to be presented here.