ACADEMIA

Complex problem solving calls for complex computing. One solution has been to make use of distributed yet connected computers in a Grid environment. And to make best use of this environment researchers are simplifying ways for users to build workflow applications. The Grid makes use of the unused processor cycles available in research networks worldwide, tapping the resources of distributed computing and data storage to provide users with seamless access to vast IT capabilities. However, the Grid itself is composed of partially cooperating and partially competing systems, and its applications grow ever more complex. In response the IST-funded K-Wf Grid project addresses the need to develop a better infrastructure for future Grid services. The project aims in particular to develop a rule-based expert system which will underpin the use of a workflow approach to solving problems, rather than the present Grid system, which is based mostly on a series of independently-running batch processing jobs. The K-Wf Grid expert system is designed to help users construct workflow-based applications. Now halfway, the project is due to finish in February 2007, and the participants have already developed an experimental prototype of the system. “A user may ask, ‘I want to produce a flood forecast for Slovakia for the period of the next two weeks’,” says project coordinator Steffen Unger of Fraunhofer First in Berlin. “This is the purpose of the K-Wf Grid system, to help users with such problems by giving advice and answers in normal language.” Unger gives another example – a user wishes to find out the traffic consequences of closing a particular street in the city of Geneva. “The K-Wf Grid system will search within Grid database resources and find any related data, then produce a small workflow that will help the user find the answers to that particular question. For example it could provide some models, so that once the input data needed for those models is entered, this will be sufficient to produce a result from the data.” The project partners are developing an ontological approach to knowledge representation for the expert system to help with the construction of workflows. Another planned innovation will be the development of a distributed knowledge-base, for storing information about the Grid architecture and the functions necessary for automating workflow construction. The three main elements of K-Wf Grid, the workflow component, the expert system and an inbuilt Grid monitoring system, will all have stand-alone capabilities, says Unger, so that each of these elements will be available for use separately. K-Wf Grid should help to bring the benefits of a global computation and information environment to a broad user space beyond that of the computer-science community. The experimental prototype is being piloted at six Grid sites, and the project partners aim to have fully working prototypes, along with demonstration applications, available by August 2006.