Skip to Main Content
The application of Grid technology is at the transition from engineering and natural science-related industrial sectors to other industries that have a high demand for computing resources. However, the diffusion of Grid technology within industrial sectors which are not naturally engineering and natural science-related is often hindered by a lack of incentives to share the computational resources. A promising way to overcome these barriers is the introduction of economically inspired mechanisms for the use of Grid-based resources. Our work introduces a iterated cost-based multi-unit resource auction (CMRA) and compares a traditional cost-based accounting approach with dedicated servers as well as a pooling approach with regard to service quality and total costs. The cost-calculus used in our model is based on costs for the delayed processing of jobs and costs for the cancellation of these jobs if the job cannot be provided at a certain time span in the worst case. The simulation results indicate that pooling of IT resources by Grid technology can produce a reduction of 20.3% in cost within this model compared to dedicated servers in the computing centers. However, with the CMRA-based allocation of computing resources, a further 1.4% of cost reduction can be achieved while the achieved quality-of-service (QoS) can be significantly increased. Finally we think that there must be a further cost reduction potential for Grid solutions beyond these savings that can be achieved by using economically inspired allocation methods that are combined with advanced refining and learning methods.