Extending Skel to Support the Development and Optimization of Next Generation I/O Systems

摘要: As the memory and storage hierarchy get deeper more complex, it is important to have new benchmarks evaluation tools that allow us explore emerging middleware solutions use this hierarchy. Skel a tool aimed at automating refining process of studying HPC I/O performance. It works by generating application kernel/benchmarks as determined domain-specific model. This paper provides some techniques for extending address situations answer research questions. For example, we document cases diverse using troubleshoot performance issues remote users, an system model, facilitating development testing mechanism runtime monitoring analytics. We also discuss data oriented extensions support study compression Exascale scientific management.