Proceedings of the 26 th Workshop of the UK Planning and Scheduling Special Interest Group PlanSIG 2007 December 17 - 18 , 2007

Service level agreements (SLAs) are powerful instruments for describing all obligations and expectations in a business relationship. It is of focal importance for deploying Grid technology to commercial applications. The EC-funded project HPC4U (Highly Predictable Clusters for Internet Grids) aimed at introducing SLA-awareness in local resource management systems, while the EC-funded project AssessGrid introduced the notion of risk, which is associated with every business contract. This paper highlights the concept of planning based resource management and describes the SLA-aware scheduler developed and used in these projects. Introduction In the academic domain Grid computing is well known, if not even established. Researchers are using Grid middleware systems like Unicore or the Globus Toolkit to create virtual organizations, dynamically sharing the transparent access to distributed resources. Grid computing started under the solely technical question of how to provide access to distributed high performance compute resources. Thanks to numberless projects and initiatives, funded by national and international bodies worldwide, Grid systems have significantly evolved meanwhile, making Grid technology adoptable in a large variety of usage scenarios. Companies like IBM, Hewlett Packard, and Microsoft have recognized the potential of Grid Computing already in the early days of the Grid development, providing noticeable efforts on research and the support of research communities. However, the Grid did not really enter the commercial domain until the present day. Already in 2003 the European Commission (EC) convened a group of experts to clarify the ∗This work has been partially supported by the EU within the 6th Framework Programme under contract IST-031772 ”Advanced Risk Assessment and Management for Trustable Grids” (AssessGrid) and IST-511531 ”Highly Predictable Cluster for InternetGrids” (HPC4U) demands of future Grid systems and which properties and capabilities are missing in current existing Grid infrastructures. Their work resulted in the idea of the Next Generation Grid (NGG) (Priol & Snelling 2003; Jeffery (edt.) 2004; De Roure (edt.) 2006). This work clearly identified that guaranteed provision of reliability, transparency, and Quality of Service (QoS) is an important demand for successfully commercialize future Grid systems. In particular, commercial users will not use a Grid system for computing business critical jobs, if it is operating on the best-effort approach only. In this context, a Service Level Agreement (SLA) is a powerful instrument for describing all expectations and obligations in the business relationship between service consumer and service provider (Sahai et al. 2002). Such an SLA specifies the QoS requirement profile of a job. At the Grid middleware layer many research activities already focus on integrating SLA functionality. The EC-funded project BeInGrid (Business Experiments in Grid (BeInGrid), EU-funded Project ) aims at fostering the commercial uptake of the Grid. BeInGrid encompasses numerous business experiments, where Grid technology is to be introduced to specific business domains. Successful experiments reached the goal of proving the benefit of applying Grid technology for commercial customers. According to the NGG, a major objective in these BeInGrid experiments is the provision of reliability as contractually expressed in negotiated SLAs. Current resource management systems (RMS) are working on the best-effort approach, not giving any guarantees on job completion to the user. Since these RMS are offering their resources to Grid systems, Grid middleware has only limited means in fulfilling all terms of negotiated SLAs. For closing this gap between the requirements of SLAenabled Grid middleware and the capabilities of RMS, HPC4U (Highly Predictable Cluster for Internet-Grids (HPC4U) ) started working on an SLA-aware RMS, utilizing the mechanisms of process-, storageand networksubsystems for realizing application-transparent fault toler-