ARGUGRID has developed a new model for building Grid-based applications through the use
of multi-agent technologies and argumentation logic to support the formation of dynamic virtual organizations.
The key advantage of the approach is simplifying the interactions between service providers and service consumers
in a Service-oriented setting supported by P2P technologies.
Over the past few years, the service-oriented computing paradigm has been advocated as a new model for developing Grid-based applications. The computational paradigm builds on the notion of a remote service which provides a uniform and standardised representation of any remotely accessible resources, e.g. software tools, computational resources, storage resources or sensor devices. Such services can be advertised by service providers, discovered by service requestors, and accessed remotely via well-defined, standards-based, interfaces.
Within ARGUGRID, our goal is to provide a new model for the provision and usage of such services based on argumentative agent technology. This is driven by a vision where intelligent agents are associated with service providers and service requestors, and where such agents can enter into negotiation and argumentation reasoning activities for service selection and composition.
Workflow systems play an important role within a service-oriented paradigm. They provide the languages and execution mechanisms that enable users to orchestrate the execution of available services and to compose them into applications. Semantic web technology can be used to help workflow users discover and locate new services that they can employ, by providing mechanisms for describing and publishing the properties of existing services using controlled vocabularies and domain ontologies.
Within ARGUGRID, our goal is to develop and use semantic workflow technology allowing service requestors to describe semantically the desired properties of the services they require for their workflows. Such descriptions are then submitted to the agents who select the real (concrete) services that satisfy the input request, as well as negotiate their parameters.
With the support of the argumentation processes, intelligent agents in ARGUGRID decide which services can be used for fulfilling the demands of users by creating, managing and joining Virtual Organisations. We view a Virtual Organisation (VO) simply as a mix of organisations and individuals who bind themselves to one another by norms or rules in order to share resources. A VO in ARGUGRID can be created dynamically within the context of an artificial agent society. Such a society provides the breeding environment of VOs and consists of:
Within the ARGUGRID framework, agents representing Grid service providers/requestors and Grid services themselves, are distributed among participating components.
- the possible agents that can be participants in a VO,
- the roles these agents can enact within the society,
- the services that can emerge as a results of agents enacting roles,
- the possible combinations of services that can be supported by the society as social activities called workflows, and
- the possible agreements agents can make amongst them, which are captured by electronic contracts to regulate interactions within a VO.
Agents and services can dynamically join and leave the system, introducing a dynamic environment which will benefit from a Peer to Peer (P2P) networking framework. This P2P overlay networking framework is a core component of the ARGUGRID middleware that operates as an enhancement to an Agent Information Service, providing scalable discovery of available agents and resources within a multi-agent framework, where a number of Virtual Organizations (VOs) are established and operate to resolve some common goals within each VO. It supports multi-attribute and range queries and improves on existing distributed K-D tree middleware solutions by implementing a new load-balancing scheme that guarantees that the query operations resolve in a logarithmic number of hops even when the load distribution among peer changes at run-time.
Argumentation, simply stated, focuses on interactions where parties plead for and against some conclusion. It provides a powerful logic-based framework for interacting agents making decision, assessing the validity of information they become aware of, or resolving conflicts and differences of opinion. It is an essential ingredient of inter-agent dialogue, negotiation, persuasion and collaborative decision making.
Argumentation can thus serve as:
Within ARGUGRID, it is the role of intelligent agents (using argumentation) to negotiate, on behalf of service requestors and service providers, which services can be used as well as an implementation plan that matches the requirements to the capabilities of the existing services. Agents are used as abstractions of users of Grid services and Grid service providers. Agents use argumentation to reason and take decisions, to negotiate with one another in order to agree on shared and coordinated access to resources and services. Agents are also responsible for creating Virtual Organisations and for the definition of workflows and contracts characterising these Virtual Organisations, derived at run-time from partial specifications provided from the different parties involved in them. Finally, agents can contribute to the resolution of disputes and disagreements with respect to contracts.
- a unifying medium to provide a model for agent-based semantic Grid systems, in that it can support the reasoning and decision-making process of agents controlling/requiring access to resources/services in the Grid,
- the inter-agent negotiation process to reach an agreement on shared and coordinated access to resources and services meeting requirements,
- the definition of workflows and contracts emerging from the negotiation and representing at a high level the resource and service composition, derived at run-time from partial specifications provided by users and service providers,
- the resolution of disputes and disagreements amongst agents with respect to agreed workflows and contracts, and, finally,
- the identification of (support for) useful information about the trust level of agents.