DOC: Include instructions on how to access the datasets used for benchmarks

[mgrover1]

To ensure reproducibility, instructions should be added on how to access datasets exclusively on LLNL resources, as with this example

https://github.com/xCDAT/xcdat-validation/blob/main/scripts/performance-benchmarks/perf_benchmark.py

This prevents full reproducibility of the results

Related to openjournals/joss-reviews#6426

[tomvothecoder]

Hey @pochedls, are all of these datasets available on ESGF? If they are, I'll check if Globus links are available to make it easier to download the larger datasets.

Also are the XML files publicly available somewhere so that non-LLNL users can open multi-file datasets with cdms2? Otherwise, we need to figure out another way to open multi-file datasets using cdms2.

[pochedls]

Hey @pochedls, are all of these datasets available on ESGF? If they are, I'll check if Globus links are available to make it easier to download the larger datasets.

Also are the XML files publicly available somewhere so that non-LLNL users can open multi-file datasets with cdms2? Otherwise, we need to figure out another way to open multi-file datasets using cdms2.

These datasets are all on ESGF.

XML files are usually created locally (to map to your local files on disk). Once the data is in place, you can create xml files with cdscan -x file.xml /path/to/dataset/ (pretty sure that is the syntax).

[tomvothecoder]

Thanks @pochedls. I'll need to update the instructions for the script so that non-LLNL users can reproduce the XMLs.