Influences between performance based scheduling and Service Level AgreementsSeptember 15th, 2011
FiledSeptember 15th, 2011A proper resource-to-job matching is of paramount importance for a better exploita- tion of e-Science environments where heterogeneous resources are shared for coordi- nated problem solving in multi-institutional virtual organizations [1]. In addition, specific requirements are often associated with compute intensive scientific jobs, e.g., weather prediction WRF1, or molecular dynamics GROMACS2, which may lead to further efficiency issues. In such computation intensive applications, a better re- sources-to-job matching can lead to significant improvements in the computation speed [2]. A performance aware job execution can be realized if there is adequate information available regarding the resource capabilities and the qualities of the ser- vices provided over the resources. A generally accepted method to evaluate and com- pare the performance of computer platforms is through benchmarking and bench- marks based metrics [3] [4].