Implementation of distributed workflow management systems

Abstract

In this work, workflow management systems are investigated and a workflow engine which can be used in the competitive and dynamic business environment for managing distributed workflows is proposed. The currently used workflow management systems’ insufficiencies revealed the need for the implementation of distributed workflow management systems. Workflow management systems ease the execution of business processes and decrease the processing time but if WfMSs are run on single servers, the workloads increase and servers become overloaded causing the operating time to be increased. Therefore multi servers are preferred for the management of workflows but the workflows have to be defined in each server separately and the definitions need to be kept updated. As the servers have to be kept updated, synchronization operations have to be scheduled frequently which increases the communication costs. Also these servers have to be using the same applications and same language for communication and execution. Another drawback of the current systems is the load of the data that are transferred between the processing systems. The job is completely transferred to the related workflow agent, causing the workload of the server increase by occupying the network. Transferring the whole job causes security vulnerability and increases the risk of loosing the data. The proposed WF engine solves these shortages. In this work, we distribute the workflow definitions between several workflow engines and send only the required data to related WF engines for processing. Using several different WF engines in the proposed system; the workload on the servers decrease, the systems become more robust and scalable, and their performances increase. Flexibility of the systems increase also since the change of related WF definition is enough for the change of the whole WF. As an additional benefit, reference of the data and state of the job that are sent to the other engines are saved in business and instance data stores in the proposed system, supplying data and state persistency in the system.