Biz2Grid

Business applications in the grid
An application for the automotive industry

 

The Biz2Grid consortium consists of the Karlsruhe Research Center for Information Technology, the Philipps University of Marburg, the University of Karlsruhe, the BMW Group and iSILOG GmbH (both of which are associated partners), as well as IBM Deutschland Research & Development GmbH (project lead). The project is set to run from 01.07.2007 to 30.06.2010. The core objective of the Biz2Grid project is to build the foundations for the use of grid technology in companies.

Within the context of industry, a migration of commercial applications to existing grid middleware is only possible if the organisational guidelines of IT infrastructures used for commercial purposes are taken into consideration. This includes standardising the corresponding processes to a certain extent, budgeting for IT costs, licensing applications and services, as well as the allocation of costs incurred to the individual cost centres.

In order to meet these requirements, existing grid infrastructures have to be adapted to the general organisational conditions and expanded through economically useful billing and pricing mechanisms.
The billing and pricing infrastructure envisaged in this project will be realised on the basis of a market in the form of a metascheduler that distributes the available resources to the service requests. The component will be developed on the basis of standardised communication protocols, thereby making it compatible with the basic D-Grid infrastructure. Biz2Grid provides a price-based scheduler for the D-Grid community and related projects. The effectiveness of the D-Grid expansion will be demonstrated using two commercial scenarios from the automotive sector. These example scenarios are to serve as the basis for best practices for future business grid solutions.

From an industrial perspective, it is necessary to consider technical aspects as well as the organisational integration of future business grid solutions into existing business processes and the operation of these new solutions as part of the overall IT landscape of a company. IBM offers a service management platform for this purpose, known as the Tivoli Service Automation Manager (TSAM). This is also to be used to address grid services and their management.

In terms of research, business models will be developed in Biz2Grid that make the worthwhile nature of participation in the grid plain to see for all stakeholders. Furthermore, Biz2Grid addresses highly technical issues. At present, the seamless gridification of applications is difficult to achieve. Thus, the challenges facing the project are represented by the distribution and parallelisation of the industry applications in addition to the design and realisation of the economic business model.
The results achieved so far in the Biz2Grid project include the following:

  • Identification and description of industrial application scenarios

  • Creation and use of a criteria catalogue for the generic analysis of application scenarios

  • Description of business models for grid participants

  • Elimination of organisational barriers

  • Prototype implementation of a market-based metascheduler Implementation of a web-based visualisation component

  • Definition of an example scenario for the dynamic provision of virtualised grid resources.


During the course of the project, these initial results are to be successively built upon and evaluated on the basis of the application scenarios identified.

Figure 1 illustrates the probable technical architecture and the interplay between the middleware, the metascheduler and the application.

If necessary, a job submitted by a user can be given a signature and encrypted before it is sent to the GridWay metascheduler. The Meet2Market component that is integrated with GridWay then uses economic indicators to decide which of the computer pools is to be used for the calculation. These pools can include grid resources (e.g. a classical cluster) as well as (Windows) desktop grids. After Meet2Market has made this decision, GridWay transfers the job to the local scheduler (either a GT4 head node or omnivore, a self-organising, decentralised scheduler that can be used in the grid).

In the desktop grid, the omnivore program is executed in each node that is involved in the calculation. The omnivore program connects to the surrounding nodes and automatically distributes new jobs to all available nodes.

If a job is sent to the GT4 head node, this forwards the incoming task to the local scheduler, the Grid Engine from Sun Microsystems (SGE). An enhancement of the SGE developed by the University of Marburg, known as the Xen Grid Engine (XGE), makes it possible to schedule jobs in virtual machines. By using modern hardware, it is also possible to run virtual Windows machines on the individual computing nodes so that Windows software (such as emPlant) can be executed on computing nodes that usually run with UNIX or Linux.