SLA-BASED JOB SUBMISSION AND SCHEDULING WITH THE GLOBUS TOOLKIT 4

High performance computing is nowadays mostly performed in a best effortfashion. This is surprising as the closely related topic of grid computing, whichdeals with the federation of resources from multiple domains in order to supportlarge jobs, and cloud computing, which promises seemingly infinite amounts ofcompute and storage, both offer quality of service (QoS), albeit in different ways.Long-term service level agreements (SLAs), which require the establishment ofSLAs long in advance of their actual usage, seem a promising way for the offeringof QoS guarantees in an HPC environment in a way that is not disruptive to thebusiness models employed today. This work uses the long-term SLA approachas a basis for the provisioning of service levels for HPC resources and presentsan SLA management framework to support this. Flexibility is provided byproviding SLAs with different service levels, support for which is integratedinto job submission and scheduling. The SLA management framework can, ona high level, be used in a generic fashion and an implementation is presentedthat is evaluated against a motivating scenario.