Keywords

1 Introduction

Due to the influence of information technology, globalization and liberalization of the economy, today’s enterprises are faced with unprecedented challenges and changes. In this new economic era, no single company can meet the market demand by their own forces. In order to occupy more market shares, a new mode of economic operation named e-commerce communities came into being. The so-called e-commerce community, basically, can be seen as a group of enterprises based on common strategies, objectives and interests of the formation of the collection. Businesses and individuals in the community are responsible for a specific mechanism, and benefit themselves from businesses and individuals as well. Through the exchange of value sharing mechanism, the basis of enterprise or individual cooperation and community management has been formed. The essence in the trends of community cooperation is structured and spontaneous, which may come from the industrial cooperation, or from the face of the common market segment and consumer needs.

In such e-commerce business model, challenges have been brought to information sharing and process integration. However, as a result, enterprise information system integration and collaboration have been motivated [1].

Application system integration is the final rendering of the business process and data flow in the information system. During application integration, issues from the internal logic of the application system changes, interactions between all applications and application data format transforming between systems are often being considered. And commonly, under the process of enterprise development, application systems are varied constantly due to the development, import and modification of different business needs [2, 3].

Traditional application integration is usually invoked by Remote Procedure Call, and Shared Data Files, or the database. However, these modes are far from satisfaction in the performance of efficiency, integrity and system maintenance. When the system has some problem, it is difficult to find out the cause. Sometimes the only solution is to modify, supplement or delete the wrong records through artificial methods. As a result, not only should the technical staff consider the business logic needs at the time of writing a program, one must also take into account operational errors and necessary processing when exceptions happened (in most applications, the part, which handles business logic, is far less than half of all orders).

Using Agent Technologies to address the inadequacy of the traditional methods are hotspots of application integration researching. Agent, generally speaking, is a software entity that completes tasks by active service. It has the features of autonomy, perception, initiative and interactivity, which equals to the general sense of giving humanistic elements to objects. These characteristics have brought new vitality to distributed computing in the Internet environments, such as people cooperate over the network, Network intelligence, and so on. Agent’s autonomy enables object service requests more transparent; allowing service requester and interoperability of middleware need only know the service interface. And under the direction of certain conditions, using Agent with active senses to provide services to request, can improve the efficiency of the service and reduce service request expenses [4, 5].

In the process of system integration, when Agent used as a single entity of multiple entities, it always has features of autonomy, sociality, responsiveness and spontaneous, which can give issues, including application integration for heterogeneous application environments required by cooperation communication, effective information sharing service, and personalized service, better implementation solutions.

This paper will discuss application system integration design based on Agent. Agent, as multi-layer entities in the system, mainly accomplishes the top services and the underlying communication, which is sorts of response services or consultation services.

2 Traditional Application Integration Models

Traditional application integration strategies are as following:

Use application program interface Application Programming Interface (API). Problems frequently encountered by using API, is that when the number of system continuously increases, API will follow the exponential increase. And data exchanged through API technology must have a fixed format, for this strategy does not support access control and error recovery features.

Pass fixed text fields through a network shared directory. Because it uses batch processing, application of this technology often leads to delays. In addition, it is usually a lack of a management mechanism for sensitive data, which is stored in a shared direction as other data files. Obviously, this technique is unsuitable for online operating environment.

Open the database to other applications. This technology solves the problem of parallel access to data, but also leads to some security concerns because users can’t access data at the same time.

Because of the improper of traditional integration technologies, our research on application integration presents opportunities and challenges. The following part will introduce the research of enterprise application based on Agent Technology [6].

3 Application Integration Architecture Based on Agent

3.1 System Architecture

We want to use AGNET technology to solve the shortcomings brought by the traditional information technology. Their logical systems are shown in Fig. 45.1:

Fig. 45.1
figure 1

Integration architecture based on agent application

Full size image

Application systems integrate the content of the application system by integrating the portal Application Integration Portal (AIP). It is worth to noting that AIP organizes the content of an application system according to the relationship of enterprise application rather than simply save all the content in the local. For example, workflow A and workflow B, which belongs to system A and system B respectively, can be used as workflows on a single system after integrating the portal AIP. And this seamless integration of the system is important for the construction of reconfigurable components.

As shown in Fig. 45.1, there is a multiple-level Agent application entity in the system and levels are classified by the application features, such as the business level, consultation level, etc. Detailed Agent level design will be introduced in the next section.

3.2 Single-Level Agent Logical Model

Single-level Agent Logical hierarchical model is shown in Fig. 45.1. In the process of application system integration, the main function of negotiation level of Agent is to coordinate activities between different applications. When an application requires additional service from another Agent, it does not access the related Agent directly, but acquire the appropriate service through the integrated Agent between enterprises.

The service level of Agent provides services for application system and Agent consultation level. Services, including user services, business services, and data services are provided. The user service function provides information, functionality and browsers to locate the user interface, ensuring consistency and integrity. Business service provides a shared business policy, generating business information from the data generation to ensure the consistency of information. It also provides the definitions of business workflow activity rule, ensuring the application system internal workflow activity restriction mechanism. Data service: Data definition, data storage and retrieval, ensure the consistency of the data. Using Agent services, application requirements can be broken down into clearly defined service. After defining the service, we need to further create specific physical components to achieve them. And these components are always universal, open interface standards and compliance, so they can be reused, and can be shared within the application as well.

Agent Middleware interface layer provides different solutions to the system integration problems. The best choice for legacy applications and open system integration is the message-oriented middleware (MOM) and ORB. The main reason to choose MOM is that it supports a number of different communication mechanisms, with which developers can break the limitation brought by legacy systems in a wide range of methods. And the object-oriented middleware (CORBA or COM/DCOM) presents a strong potential in the Integration of legacy systems as well. Having advantages like advanced software bus and object-oriented technology, CORBA is easy to achieve the integration of legacy systems. It is also allowed a packaged legacy application to be reused as a component of a structure based part.

Web services also implemented the mechanisms of service requesting, service registration and service delivery, which greatly simplify the integration with distributed legacy systems.

4 Service-Oriented Working Mechanisms Based on Agent Integration System

Web service is the mainstream of distributed computing, so the core of Integration system based on Agent is web service. The framework of application integration using Agent is shown in Fig. 45.2. In the framework, application system is central on the service, mainly including the following contents:

Fig. 45.2
figure 2

Application of integration working system model based on Agent

Full size image

Master Matching Agent: the entrance of the application integration system for service requester and implements the master matching function.

Provider Matching Agent: Packaged in each web service provider system, implement the detail matching for each web service request.

Master Service Registry Repository: Store all the information of registered web services in the entire application integration system, and provide web service matches for MMA. The storage of web service is according to the web service description language basing on XML format.

Provider Service Registry Repository: Store the complete register information of web service for providers, and offer detailed matches for PMA. The register information is stored according to the web service description language.

In this application integration model, a service request can finish a series of service matching operations through integrating the platform AIP. The working mechanisms are shown in Fig. 45.2:

  1. (1)

    Service requests submitted a web service through the client to a WEB page in the server;

  2. (2)

    Agent Manager will stimulate a Session Agent (SA) according to the service request;

  3. (3)

    When the SA knows where the demand comes from, it will return the user a Client-Applet, through which guide the user to fill in the basic web service query information;

  4. (4)

    Pass the information to SA again;

  5. (5)

    SA will extract the necessary data and convert it to XML format query description. Then, this information will be passed to MMA;

  6. (6)

    MMA will automatically search the matching web service set in the related MSRR according to the description. And return the result set to SA;

  7. (7)

    SA returns the results to users;

  8. (8)

    Users on the browser side will select one or more web service providers depending on the result set and then submit again;

  9. (9)

    After receiving the specific web service provider information, SA will connect the PMA directly, and implement web service dynamic binding at the same time;

  10. (10)

    The appropriate provider’s PMA will match the web service request according to PSRR. And the proper service system will establish a connection to the service request through SA in the end.

5 Conclusion

Along with the research on the application of artificial and distributed collaborative computing, people have discovered that Agent technology can offer a great solution to address the issue of coordination and integration, which is due to the advantages of Agent system, such as autonomy, self-adaptability. In multi-level Agent system, the restrictions on centralized control, planning control and sequence control are released, and new functions, including distributed control, emergency and parallel processing are added. Another advantage of Agent system is that it can provide efficient service with low cost. From the discussion, we may acknowledge that Agent technology is a promising technology, and we hope that the research in this paper can offer help to the related researches and applications.