هند: چارچوبی برای حمایت از ایجاد قابلیت همکاری جریان کار به معنای سیستم های چند عامله

Inter-organisational business processes are by no means rare. On the contrary, a business can only thrive if it cooperates with other businesses, like service providers, wholesalers, shipping companies, insurances, banks, etc. However, there is a major barrier here for automating such processes: the IT infrastructure of the trading partners, or more specifically their variety and independence. Owing to these problems, even if both organisations have automated support for their parts of the whole inter-organisational process, the activities responsible for cross-organisational messaging or document passing are often done manually by post or email. This paper concentrates on the problem of providing support for workflow interoperability, which would allow for linking business partners without the necessity of a major redesign of their workflows. A framework for interoperability between heterogeneous workflow systems is presented, which uses a layer of agents to enforce the correct execution of coordination dialogues between the business partners. Further, the framework is discussed and conclusions are given.

Autonomic computing is described as the ability of an IT infrastructure to sense, respond and ultimately adapt to change in accordance with business policies and objectives. The aim is to free IT professionals to focus on higher-value tasks by making technology work smarter, with business rules guiding systems to be self-configuring, self healing, self-optimising, and self-protecting (IBM 2001; Kephart and Chess, 2003). As such autonomic computing presents a grand challenge which may only be solved by combined efforts of researchers from disparate domains. One domain which can both contribute and greatly benefit from autonomic computing is the domain of workflow technology. Over the past three decades workflow technology has become established as one of the backbone IT infrastructures of many large organisations. Early work concentrated on supporting the flow of work within small organisational units. It has then gradually expanded its reach to the departmental, and later, to the corporate level. In the current economic climate, with the emergence of businesses having world-wide presence, with companies merging, being taken-over, forming alliances, etc., the integration of business processes plays a crucial role. While some may use standardised workflow products, many still use non-standard or in-house solutions. This situation has prompted a shift in workflow research, encouraging researchers to start looking into ways of applying workflow technologies to cross-organisational processes. Providing interoperability solutions brings a number of benefits to the organisations. They include: savings on costs of development/implementation of new workflow processes, savings on time spent on staff training, and respect for work culture (existing users are not deprived of using well-known, pre-existing tools). Recognition of the importance of workflow has led to the formation of several non-commercial bodies that have endeavoured to standardise workflow systems. The founding of the workflow management coalition (WfMC) in 1993 started the process of developing workflow standards beginning with the workflow reference model and its five interfaces (WfMC, 1995). Later other organisations looked into workflow-related issues: it is worth mentioning the work done by the object management coalition (OMG) on adapting WfMCs workflow facility to the object-oriented context. The standards cover a wide range of workflow issues starting from process definition, through its execution, monitoring, and interoperation with other standard workflow systems. The problem is that the standards cover simple workflow interoperability models, mainly chained or nested sub-process execution. However in many cases cooperating companies require more complicated, parallel-synchronised workflow execution. This is usually implemented on a case by case basis. Two organisations first agree on a specific exchange of data and/or messages, and then each one of them builds their workflow independently. This approach has a large potential for errors. When each organisation builds their workflow separately, problems frequently occur when systems are connected, much time is spent on testing, and very large-scale linking interfaces may become unmanageable. Work done in the Kanagawa Institute (Kanagawa Institute, 1999) aimed to rectify some of the problems by first allowing the semi-automatic generation of workflows from an agreed upon inter-workflow definition. The next stage is to let the companies fine-tune the generated workflows by adding the internal details of their operation. This framework was a step in the right direction. The meta-definition of workflow could be considered a cooperation scenario between two partners, which might be reapplied in similar situations. However, this solution has several disadvantages. If companies have already invested in the development of their workflows it will not be financially viable to build them again from the beginning. If a change is needed in the inter-workflow definition both workflows would be wholly re-generated. Finally, the framework has to be able to create process definitions understood by the workflow systems of various vendors, which is a problem with all frameworks aiming to be generic and is usually rectified by using vendor-specific plug-in components.

This paper commenced with an observation that current workflow systems do not cover the issue of interoperability sufficiently. The lack of standardisation, especially in the area of parallel-synchronised processes, forces the developers to build custom solutions, tailored for the specific needs, products, and systems. Yet, companies face the pressure to build value chains with many business partners, which raises the cost of managing and developing such custom solutions to the level unacceptable by most small or medium enterprises. On the other hand, companies have invested in the implementation of workflow systems and are not inclined to discard them. The work presented here is an attempt to resolve the mismatch between various business partners, which exists in the area of workflow interoperability. Autonomic computing is a relatively new (IBM, 2001) trend, initiated in by IBM only several years ago. However, despite its youthful stage, it is already well established on basis of previous research and commercial work in the fields of agent technologies, inter-networking, workflow to name but a few. Our approach is relevant in this context. In InDiA, dialogue agents autonomously initiate and control the progress of cross-organisational workflows thus freeing the IT professionals from the tedious details of low-level inter-workflow communication allowing them to concentrate on the relatively higher level specifics of message exchange. For now, it can be concluded that InDiA is a feasible approach to workflow interoperability. It manages to achieve it while leveraging the existing workflow technology, supported by a layer of agents which proves to be a convenient intermediary amenable to further extension as the workflow technology evolves. The extension should aim at a greater autonomy of the agents in the context of disparate computing systems and dynamic changes of business partners. The next step is to enhance the agents with the ability to better handle failed workflows with the view of switching business partners when the required quality of service is not met.