Talk Professor Volker Sander

Professor Volker Sander will visit the UvA on September 28 and 29 to initiate a collaboration with e-science and workflow group (sub-group of SCS).

Biography:

Prof. Volker Sander, Dean of the department Medical Technology and Technomathematics of FH Aachen University of Applied Sciences. He has been actively researching issues in eScience since 1996 as a research scientist at the Jülich Supercomputing Centre of the Research Centre Jülich. In 1999 and 2000 he was a visiting scientist at Argonne National Laboratory where he was as a member of the Globus project responsible for prototyping a next generation resource management system for the Globus Toolkit. Dr. Sander was the spokeman of the Grid Computing domain of the Research Centre for the program oriented funding in 2004. Dr. Sander has been involved in Grid standardization efforts since the very beginning and was part of the Internet2 QBone Signaling Design Team. He was acting as co-chair of the Grid Resource Allocation Agreement Protocol (GRAAP) Working group of the Open Grid Forum (OGF)and is co-author of multiple OGF, and an author and co-author of various scientific conference and journal papers in this area. Now he is responsible for the degree program of more than 800 students including a Bachelor study in Scientific Programming held in collaboration with the Research Centre Jülich and RWTH Aachen and a Master study in Technomathematics/Scientific Computing held in collaboration with the Research Centre Jülich

Abstract:  Current e-Science infrastructures provide support for complex scientific processes that consist of orchestrated resources such as pure computational devices, specific applications, data repositories, or scientific instruments. Grid environments are the common technical approach used to build these e-Science infrastructures on a middleware platform that provides the essential services. In order to support the orchestration of scientific tasks, many Grid middleware platforms offer a workflow management system either as an integrative part of their middleware distribution or as an enactment service build on top of the middleware. To date, a variety of popular Grid workflow management system exist that use an approach that follows the push-pattern of the known workflow resource patterns. Here, a software agent, e.g. the workflow engine, actively exercises control about the progress of a workflow by pushing the individual tasks to the selected resources according to the dependencies, provided by the workflow description. The reason for the popularity of this pattern is that the workflow engine can here be viewed as an enactment service that can be dealt with like with other services, such as a scheduling service or a data transfer service. Clients can create their workflows and submit them to the workflow management system by standard Grid mechanisms. Of course, the user acceptance of the provided service depends on the quality of the provided service. To improve this quality, various advances to the management system have been performed to assure data or process oriented provenance or to allow for a late binding of tasks to resources. Particularly for the latter, a fundamental barrier remains with respect to the acceptance: The push-model requires that a service receives a particular level of control about the resources. While this barrier is addressed by the foundation of a virtual organization (VO), by a stringent management of the VO and by the introduction of Service Level Agreements, a pull-based approach could offer various new perspectives for e-Science infrastructures. Here, resources can not only actively select which work item they wish to commence next, it also allows the integration of human interaction patterns according to emerging standards such as WS-HumanTask. This talk presents an idea for a pull-based approach for workflow management in Grid environments. The approach was motivated by a specific proven ance concept in a single e-Government application domain that relies on layered, digitally signed XML-documents that traces the history of operations in a legally usable form. It generalizes the idea of using layered, digitally signed XML-documents to trace the progress of a workflow and introduces a pull-based approach to Grid workflow management by an intermediary that mediates between workflow instances and resources.