Case studies

[khou2020]

Propose case studies

[wkliao]

@khou2020 @yzanhua
Please check h5bench and see if the I/O patterns there can be case studies.
https://h5bench.readthedocs.io/en/latest/

[khou2020]

@khou2020 @yzanhua Please check h5bench and see if the I/O patterns there can be case studies.

They are similar to what we have in terms of I/O pattern:
h5bench read/write: a selection of block appending, strided (interleaved) I/O pattern that can be recreated with IOR.
Metadata Stress: Subarray pattern but written in many rounds
AMReX: Similar to FLASH
Exerciser: Subarray pattern
OpenPMD: Similar to FLASH
E3SM-IO: Borrowed from our benchmark

The differences between them are only in file metadata (attrs, #vars, var dimensions ...) and a slight difference on data assignment (ex. different blocks per process between blocking appending patterns).