Concept for Adaptive Composition and Trading based on Agents (ACTAS)

Fig. 1 - The Service Composition Framework of ACTAS

In recent time, Service Composition and Enterprise Application Integration in Distributed Information Systems (DIS) became a more important subject of research. An Information System describes a principal design for wrapping the application logic in a presentation layer and resource management layer, which allows one to offer the functionality of the application as a service. A Distributed Information System is based on middleware, which facilitates and manages the interaction between applications across heterogeneous computing platforms. The dissertation introduces a framework for Service Composition as a principal extension of such a middleware for Distributed Information Systems (Fig. 1).

The idea of Application Integration is not new. Especially in E-Business, so-called Enterprise Application Integration (EAI) was introduced quite early. Middleware for Distributed Information Systems supporting EAI allows the composition of different services in order to fulfil a Service Request. Workflow management systems are a specialisation of such a middleware. With Web Services and their standardizations, Service Composition became more feasible and important for E-Business (BPEL4WS, OWL-S, EnTish). An obvious example for a Composite Service in E-Business is the Supply-Chain for a product, which composes several services offered by the supplier and the warehouse.

Fig. 2 - VPO with Working Zones

The problem of Service Composition is not unique for E-Business. The dissertation also introduces an example situated in the area of CSCW (Computer Supported Cooperative Work). The research in the area of CSCW does not only involve Computer Science, but other research disciplines like for example Sociology, Psychology, and even E-Business. The CSCW research project COR/UNITE of the Fraunhofer research institute for Secure Telecooperation (FhG-SIT) in Darmstadt describes and implements a virtual environment, the so-called Virtual Project Office (VPO) (Fig. 2), for supporting the project-based cooperation of a worldwide-dispersed group. Each member of such a project team, who is currently working at the project, is represented as an avatar in one of the working zones of the VPO standing for this project. Working zones, like for instance a zone for a conference room or a common working space, are metaphors illustrating the way the group member is presently involved in the project. A challenge of this environment is to establish a communication service, which uses the currently accessible communication facilities of the Service Clients, who want to communicate. These communication services must obey the policies connected with the current project and the working intentions of the Service Clients. A policy can be described in the profile of a Service Client and his/her current working zone in the VPO. The resulting communication service is a Composite Service with Component Services established by the involved communication facilities. Programming middleware like JINI makes it possible to offer and access technical facilities as a service. In fact, the subproject World Wide Facility Management (WWFM) of COR/UNITE is concerned with the access and support of technical facilities of buildings by so-called room-computers.

There are several challenges, which have to be tackled by a concept for Service Trading and Service Composition:

· The availability of services and their resources: Although services are available in general, they may not be when the trader wants to recommend them. Thus, the trader needs to check the current availability before recommending a composed service.

· The trading for new Component Services, which fulfil requested properties in a given context and is principally compatible with the already selected component services.

· The composition itself: The parties involved have to agree on a common language for requests and the composition model. Furthermore, the Component Services of the resulting Composite Services in COR are different in their composition from the ones in E-Business.

· The grounding of the Composite Service and its Component Services has to reserve the resources, to determine the coordination as well as the order and times for the scheduling of the Component Services.

The dissertation illustrates these aspects of Service Trading and Service Composition. It further presents a framework for composition based on a lifecycle for the Composite Service and its components. The framework is called “ACTAS – Adaptive Composition and Trading based on Agents” and it improves the way to find compatible and secure Composite Services for telecooperation and other applications (Fig. 1). ACTAS observes the availability of the services and offers a Service Model with an integrated Composition Model for the service description. There are several service description technologies like for example WSDL, DAML-S, T-Models of UDDI, and DF-Agent-Descriptions of FIPA, which among other ones are described in the state-of-the-art chapter of this work. Other description technologies will be developed in the future. Many of them developed their own kind of criteria for compatibility of services. Thus, there is a motivation to design a middleware for Service Composition as an adaptive framework for new kind of descriptions. The Service Model of ACTAS allows the definition of property classes, which are used for the declaration of properties in a (part of a) semantic context, a so-called Characteristic. With the introduction of object-oriented properties, it is possible to preserve not only the type of a property description, but also its behavioural semantic. The Service Provider describes the different modes of a service and their compatibilities through the mentioned Characteristics.

Fig. 3 – Multi-Agent System of COR with ACTAS

The UNITE/COR project uses software agents for its implementation. Since this project gave the original motivation for the hypothesis of this dissertation, the realisation of ACTAS was also based on agents (Fig. 3). It turned out that the properties of software agents are essential, in order to create a framework for Service Trading and Service Composition like ACTAS. With the use of agents for the trading and composition of services, different goals and policies for Trading and Composition can be pro-actively achieved. The management of the availability of services demands a re-active behaviour. The behavioural semantic of the Service Properties especially in the merge process of the composition is realized by the agents. Furthermore, the application of agents allows the separation of Service Trading and Service Composition from the Service Grounding and Service Running.

Intelligent Agents

Agents are being used in an increasingly wide variety of applications - ranging from

comparatively small systems such as personalised email filters to large, complex, mission
critical systems such as air-traffic control.

However, what is an agent?

There is no definitive answer to this question. Many people see in an agent an extension of an object, which allows an agent to react to external events and to influence this environment. Autonomous agents are distinguished from servers. A comparison between Agents and Objects can be found in [1].

Some qualitative criteria of an intelligent agent are:

· Agents are capable of flexible actions in order to meet their design objectives. This flexibility
of an agent can be expressed by three terms: responsive, pro-active, and social.

Agents should perceive their environment and respond in a timely fashion to change
that occur in it (responsive)

· Agents should be able to exhibit opportunistic, goal-directed behaviour and take the
initiative where appropriate. They can even learn to improve their behaviour. (pro-active)

· Agents should be able to interact with other artificial agents and/or humans. Several agents
can form a system (multi agent system, MAS) (social)

[1] Wooldridge, Michael: Intelligent Agents in: Weiss, Gerhard (Editor): Multiagent Systems, A Modern Approach to Distributed Artificial Intelligence, MIT Press, 1999