جریانات کاری اطلاعات نام نویسی به عنوان یک مکانیسم کنترل و هماهنگی برای پردازش استثنا در سیستم های جریان کار

A higher level control and coordination mechanism is required for exception handling in workflow systems. This paper describes such a framework based on events, states, and a new kind of process called a meta workflow. Meta workflows have five kinds of meta activities and facilitate control over base workflows. We describe the framework and illustrate it with examples to show its features. The paper gives an architecture for incorporating it into existing workflows and also provides a formal semantics of execution. This framework can be used in Web services, supply chains, and inter-organizational applications where coordination requirements are complex, and flexible and adaptable workflows are needed. It is also useful for handling not just failure recovery but also various kinds of special situations, which arise frequently in web-based applications.

Web services and supply chains [3], [13], [17] and [12] are workflows that span multiple organizations. Thus, they require coordination and control of both the data flow and control flow across multiple organizations. Exceptions are a common occurrence in workflows [4], [11], [15] and [18]. Even a seemingly simple process like a travel expense claim or order processing can become difficult to describe if one tries to cover all the special situations and exceptions in the description. This creates a very awkward process description that is hard to read and understand, and also error-prone. Therefore, most workflow systems are able to capture the simpler form of a process and tend to collapse when variations are introduced. Exceptions can be planned or unplanned. A planned exception is an abnormal situation that has been anticipated and a way for handling it has been included in the process. On the other hand, an unplanned exception is one that has not been anticipated. In this paper, we present a formal methodology for describing exceptions in a workflow. An exception is a special situation that occurs infrequently in a workflow. The main idea is to describe a basic, primary process first and treat abnormal and infrequent situations separately as supporting workflows. Our goal is to provide support for the planned exceptions and also be able to incorporate the unplanned ones relatively easily. We introduce two new notions, meta workflows and ESP (Event-state-process). A meta workflow is a special, higher level control process that consists of five control commands: start, terminate, suspend, resume and wait and suspend. An ESP rule causes a meta workflow to run when an event occurs and a workflow case is in a certain state. This may cause a meta workflow to execute, and thus perform control operations like suspending certain workflows and starting other workflows, etc. Thus, we make a clear distinction between two types of workflows: base workflows and meta workflows. The base workflow corresponds to specific tasks that must be performed. The meta workflow is only for control purposes and consists of the control commands described above. In general, when it is not qualified the term workflow refers to a base workflow. The advantages of this approach are modularity, extensibility and adaptability. The basic workflow description is kept simple while variations to the basic process are described separately in a modular manner. It is possible to add new ESP rules and corresponding base- and meta workflows when a new situation arises thus giving extensibility. Minor changes may be made to ESP rules (or a module) to adapt the workflow to certain situations. Such small changes would be hard to make on a monolithic workflow. Finally, the simplicity of the approach helps in minimizing errors, while at the same time making it easier to describe complex situations. Consequently, this framework constitutes a new methodology for workflow modeling which has applications in various kinds of web services. The organization of this paper is as follows. Section 2 gives a formal description of our framework along with semantics. Then Section 3 illustrates the framework with examples. Next, Section 4 discusses an architecture and expressive power of our framework, while in Section 5 we present some discussion of this approach in the context of related work, in particular BPEL4WS. Finally, Section 6 concludes this paper.

This paper described a framework and architecture for support of exception handling in business workflows. It is based on detecting event-state combinations that cause higher level processes called meta workflows to be executed. A meta-process consists of five meta activities for controlling and coordinating base workflow processes. We demonstrated the usefulness of the approach and gave an architecture for integrating this approach into a current workflow system. The advantages of this framework are modularity, extensibility and adaptability. We foresee further research along several lines. First, we already have a prototype implementation of XRL (eXchangeable Routing Language) technology [19] and are planning to use this as an experimental test bed for adding the functionality proposed in this paper for meta workflows. Such a test bed would allow us to evaluate the proposal from a performance standpoint. Secondly, although we gave the semantics for execution of meta workflows themselves, rule conflicts were ignored since we assumed that the rules are ordered by priority and the first applicable rule fires. Future work could address this by developing detailed rule execution semantics. Thirdly, support for composite events would be a useful feature. At present each event is processed as soon as it occurs, and if no matching ESP rule is found it is discarded. Support for composite events would allow for action to be taken on multiple events that are related to one another.