Coarse Grained Parallelized Scientific Applications on a Cost Efficient Intel Atom Based Cluster

Modern science requires the ability to evaluate large numbers of results from experiments. Therefore, automated interpretation of experimental results often has a significant influence on the costs of a scientific project. One way to reduce these costs is to use ultra low cost and energy-e_cient hardware for building analysis systems. In this paper two cluster systems which make use of Intel Atom micro-ITX boards and other low cost peripherals are compared. Three real world applications are chosen to proof the concept of the systems and compare them. Additionally, synthetic benchmarks are conducted to put the performance of the applications in perspective. The results of the measurements are discussed and used to evaluate the effectiveness of the systems in terms of performance, power consumption and costs. Finally, based on the conclusions, specific application areas are suggested in which the introduced systems perform well and in which it is better to use clusters with high end server components