h5bench

h5bench is a suite of parallel I/O benchmarks or kernels representing I/O patterns that are commonly used in HDF5 applications on high performance computing systems. H5bench measures I/O performance from various aspects, including the I/O overhead, observed I/O rate, etc. You can check h5bench’s source-code at https://github.com/hpc-io/h5bench. You can check h5bench contiguous performance regression experiments in the Perlmutter supercomputer (NERSC) at https://bit.ly/NERSC-h5bench.

These are the benchmarks and kernels currently available in h5bench:

Benchmark	Name	SYNC	ASYNC	CACHE	LOG
h5bench write	h5bench_write	✅	✅	✅	✅
h5bench read	h5bench_read	✅	✅	✅	✅
Metadata Stress	h5bench_hdf5_iotest	✅	⬜	⬜	⬜
AMReX	h5bench_amrex	✅	✅	⬜	⬜
Exerciser	h5bench_exerciser	✅	⬜	⬜	⬜
OpenPMD (write)	h5bench_openpmd_write	✅	⬜	⬜	⬜
OpenPMD (read)	h5bench_openpmd_read	✅	⬜	⬜	⬜
E3SM-IO	h5bench_e3sm	✅	⬜	⬜	✅
MACSio	h5bench_macsio	✅	⬜	⬜	✅

Getting Started

• Build Instructions
  • Build with CMake (recommended)
  • Build with Spack
• Running h5bench
  • h5bench (recommended)
  • Manual Execution

Benchmarks

• Read / Write
  • Configuration
  • Supported Patterns
  • Understanding the Output
  • Known Issues
• Metadata Stress
  • Configuration
• AMReX
  • Configuration
• Exerciser
  • Configuration
  • Exerciser Basics
• OpenPMD
  • Configuration
• E3SM
  • Configuration
• MACSio
  • Configuration

Contribute

• Ways to contribute
  • Reporting bugs
  • Suggesting enhancements
  • Adding new benchmarks
  • Testing

Legal

• Copyright
• License