SZTAKI - Department of Distributed Systems

Because of its apparent strengths in distributing resources, and coordinating their activities towards an overall objective, grid computing has a high potential to contribute substantially to the competitiveness of commercial and non-commercial organizations. Most current Grid solutions, however, are still not able to sufficiently achieve this aim because grid technologies still lack the capability to cope with important challenges enterprises are faced today.

Successful enterprises focus their activities onto their core services. This leads to, often complex, supply chains involving lots of enterprises (e.g., in case of large airports up to 500 companies). These enterprises exhibit high responsiveness, i.e., they are able to meet their customers needs quickly and in high quality, and they adapt themselves successfully to changing demands of and situations in their markets without loosing much time. They perform their processes very efficiently. This requires them to manage intra- und inter-organizational interactions so that every days conflicts between antagonistic aims of different players, and conflicting access to resources can be resolved at low cost, and that each opportunity to get advantages from possible synergic effects can be addressed. Finally, successful companies are able to make good decisions even in case of incomplete, inconsistent, wrong or missing knowledge.

Till now, grid technologies do not meet these challenges sufficiently. The BREIN project thus aims at compensating these deficiencies by taking a more "business-centric" approach. It develops towards a framework, which involves formal representations of supply chain structures, business goals, skeleton process models, process planning and process instantiation patterns, patterns for intra- and inter-organizational interactions, formal representations of the aims, capabilities, and activities of humans as well as of technical (artificial) actors, and simulation toolkits supporting complex decision making in non-standard decision processes. To this end, BREIN will combine the flexibility and stability of Grid-technologies with strength of methods for business modeling, with the intelligence and adaptability of AI technologies, namely Agent theory, planning & optimization strategies and with the semantic web.

To this end, BREIN will make use of results from projects like TrustCOM or NextGrid with respect to dynamic virtual organisations, security issues etc. with an explicit focus on the participant's business needs. The expertise of the business partners from all over Europe, will ensure that an according set of requirements will be generated.

Project objectives

Research in agent technology is developing theories, software engineering frameworks, and standards for computational entities (agents) that are capable of exhibiting flexible behavior in dynamic, and unpredictable environments. They are capable of autonomous action in order to meet their design objectives. The weak notion of agency comprises four properties which are regarded as necessary, and sufficient for agent hood (Luck et al. 2004): autonomy (actions function without any direct control by programmers or user), reactiveness (adapt their behaviors quickly and effectively to changes in the environment), proactiveness (to plan and execute their actions with respect to their expectations about future changes of their environments), and social abilities (to identify other agents in their environment, and to communicate with them). The strong or intentional notion of agency also requires agents to be based around control architectures comprising mental components such as beliefs, desires, and intentions. In the context of BREIN, to address the development of successful grid applications for business, five major achievements of agent technology are relevant:

• Head-body-architectures provide an easy-to-use solution by which any grid resource can easily be encapsulated (virtualization), and at the same time integrated into a multiagent architecture that provides protocols for the coordination of decentralized activities.

• Cooperation, and coordination: A large number of sophisticated cooperation and coordination protocols provide powerful services for for distributed negotiation, cooperative planning, cooperative interpretation, and cooperative action.

• Knowledge modeling, and interpretation: Agents exhibit the capability to resolve knowledge-related concerns (e.g., uncertain, fuzzy, incomplete, inconsistent, wrong or missing knowledge) through mutual cooperation. This is in particular important in highly dynamic environments where it becomes impossible to get fully consistent snapshots of larger sections of the world of discourse.

• Situated behaviors: The formal representation of beliefs, desires, and intentions enables agents to perform their behaviors under local control, without any direct intervention of programmers or users (situated behaviors). This provides strong support to leave local concerns with the agents, and to only design global concerns into cooperation and coordination protocols.

• Group formation: Based upon their independent decision, agents within an agent society can dynamically establish (and dissolve) multiagent systems, based upon cooperation/dissolvement agreements (e.g., contracts) among some members of an agent society.

This Integrated Project proposes to build on the results achieved in the realisation of Next Generation Grids and Semantic Grids and to further enhance them by integrating research results from the Agent domain with a particular focus on approaches to form collectives of agents, even if these agents have conflicting goals or antagonistic behaviors. Apart from aiming for higher reliability and scalability through a more decentralized approach, we believe that the configuration, control and maintenance of service providers and also Virtual Organizations (VOs) can be executed on a higher abstraction layer, such that this does not necessarily require in depth knowledge of Grid technologies. In few words, we aim at the "invisible Grid", which will just appear as a powerful enabling technological infrastructure to support real fore-front business cases.

In fact, this consortium believes that the strong focus in recent Grid research to optimize the operation of Grids and hence Virtual Organizations is not in the main interest of Service Providers - instead, it is only focused on achieving the best result for an "ultimate" consumer of the VO outputs. Since VOs have been built to realise mechanisms for resource sharing where all VO participants share a common goal, such an approach was generally appropriate. However, assuming that Service Providers are offering their resources to different VOs at the same time, each Service Provider is trying to optimize the service's performance according to his/her business objectives that must not necessarily be in line with the objectives of the VOs to which these resources are offered. A service provider may for example decide to intentionally violate a Service Level Agreement in order to fulfill another SLA for a more important customer or where a higher penalty would be applied.
The main problems identified in current Grid solutions that prevent the exploitation within the business domain so far are:

• Difficult to use: The existing solutions have never been designed for application in a wide scale and for inexperienced users. Offering services or even using services offered by a Virtual Organisation or a single business entity is still difficult and cumbersome.

• Lack of Autonomy: Real life business concepts imply decision taking by the respective business participants, but current Grid solutions don't provide these mechanisms to support the user

• No goal-oriented business processing: Current Grid architectures are designed for an optimised usage of resources which means today to use as much resources as possible in order to optimise execution time or storage performance.

• Non-optimal management of distributed execution: Workflows in the current Grid architectures are static and therefore provide no base for optimising them with respect to the needs of the customer. A workflow once designed is not able to be changed to consider policies or the intentions of a customer. If a workflow foresees a specific Service Provider to do a specific task and this kind of Provider is not available, it might happen that the executions of the workflow won't take place at all.

• Non-optimal management of individual performance: Optimization of services with respect to the needs and goals of the respective Service Provider is desirable but not supported yet in today's Grid Architectures.

• Lack of appropriate Security models: Originally the participants of Virtual Organisation shared their resources and had only a common goal. So current Grid Architectures are very typically limited to Authentication and Authorization and the appropriate mapping of credentials on the used resource. This security is realised using application level means increasing the probability for failures and increasing the complexity for implementers. This means that there is currently an unacceptable gap between high those lower-level system security policies and mechanisms and the desired high-level business/individual security requirements.

• Inappropriate modelling constructs: Architectures, protocols, cooperation models, resource concepts, and VO models that are still not able to adequately reflect relevant business concerns. This requires to develop much more detailed, and sometimes more flexible definitions for the above mentioned concepts. Another issue is the lacking market/customer perspective: enterprises need to succeed at the market, i.e., to fulfil customer demands better than its all competitors. This is an important issue for enterprises requiring them to establish customer-centric value chains, to keep their own cost low, and to invest only in new (risky) technologies if these offer an appropriate chance for commercial improvements.

• Lacking contribution to competitiveness: Enterprises adopt new technologies only if these technologies offer a significant degree of competitive advantage; increases in efficiency (cost reduction), or increases in effectiveness (better or new products/servies for the customer).

BREIN focuses on the approach of fusing Grid, Agents and Semantic Web supported by the so-called business factor. BREIN takes the e-business concept developed in recent Grid research projects, namely the concept of so-called "dynamic virtual organisations" towards a more business-centric model, by enhancing the system with agent technology and methods from artificial intelligence, semantic web etc. In business-terms, BREIN has a clear vision of developing a business-oriented grid solution with integrated intelligence by combining the Grid with agent technology and semantic web.

The BREIN vision foresees a world in which:

• agent technologies as well as semantic web technologies enhance the current Grid to an intelligent knowledge Grid powerful and useful for business and society

• complexity has been significantly reduced by using data abstraction to enable interaction with the system without specific knowledge of the technology inside

• flexibility and stability of the system are increased by integrating mechanisms for self-adaptation, self-management, self-optimization, self-healing and self-organization

• the individual business needs ("high performance", "low costs") of participants or a group of participants (VO) are pursued by a decentralized Grid where decisions and actions are taken by agents with respect to the objectives of the respective participant(s)

Results

SZTAKI has developed a plug-in for SOAP communication, which enables the JADE multi-agent platforms to communicate via SOAP messages instead of HTTP or IIOP. The plug-in can be considered as a lightweight agent web service integration toolkit, so that the benefits of both technologies can be exploited simultaneously.