Scheduling for Responsive Grids

en cours de soumission chez Kluwer Academic PublishersIn the 70s, the transition from batch systems to interactive computing fueled the widespread diffusion of advances in integrated circuit technology. Grids are facing a similar challenge, namely the seamless integration of the grid power into everyday use. One critical component for this integration is responsiveness, the capacity to support on-demand computing and interactivity. A large contributor to responsiveness is the Quality of Service (QoS) for the job execution time. Grid scheduling is involved at two levels in order to provide QoS: the policy level and the implementation level. The main contributions of this paper are as follows. First, we present a detailed analysis of the performance of the EGEE grid with respect to responsiveness. Second, we define and demonstrate a virtualization scheme, which achieves QoS, schedulability analysis, and coexistence of QoS and best-effort poli- cies, within a brokering-based system organized as a federation of batch-scheduled clusters. Last, we examine two user-level schedulers located between the general scheduling layer and the application layer. These are the DIANE (DIstributed ANal- ysis Environment) framework, a general-purpose overlay system, and a specialized, embedded scheduler for gPTM3D, an interactive medical image analysis application