P57: Adaptive Tier Selection for NetCDF and HDF5
Abstract: Scientific applications on supercomputers tend to be I/O-intensive. To achieve portability and performance, data description libraries such as HDF5 and NetCDF are commonly used. Unfortunately, the libraries often default to suboptimal access patterns for reading/writing data to multi-tier distributed storage. This work explores the feasibility of adaptively selecting tiers depending on an application's I/O behavior.
Award: Best Poster Finalist (BP): no
Two-page extended abstract: pdf