| http://www.w3.org/ns/prov#value | - CAP Considerations * How much time will profiling experiments take? - Use appropriate data volumes and concurrencies - Tradeoff of science vs. tuning * Interpreting data - Some data is easy to interpret, e.g. MPI_Barrier() time - Other data requires a performance model to understand whether a measurement is good or bad * Tool Robustness - Increased data volume and concurrency increases chance of t
|
