This research focuses on providing a workflow model that is based on peer-to-peer architecture. Distributed and parallel computing is the design concept. It will provide maximum performance and give the system the highest scalability, flexibility and configurability. This model contains a supervisor-agents working unit, a hub that is a housekeeper for all data transaction, a workbench that provides a platform for all workflow configurations. Each agent is responsible for a specific task, like converting a rile from one format to another, or, getting input from end user. A supervisor can take care of multiple agents. One computer has one supervisor. Supervisors are responsible for agents' load balance, launching/killing agents, communicating with other supervisors to keep the whole system at its peak performance. If we call supervisor-agents working unit as one peer, the model consists all peers and peers are communicating each other to keep the system running Hub stores all system data, like configuration data, task loading data, etc., and all work files, like input/output, database transactions, metafiles, etc. It also has a system wise database to keep all system logs and task balancing. Workbench stands for a platform or UI for users to configure their workflows. A workflow can contain many working steps; each step calls one agent to fulfill the real tasks. Depends on the customer needs, users can have many steps as they want to. And also they can have many workflows as necessary.
