Filesystem Benchmarking suggestions are incorrect

https://docs.gitlab.com/ee/administration/operations/filesystem_benchmarking.html

The problem with this suggestion (using fio) to benchmark filesystem performance is that fio does NOT create realistic file workloads. The suggestion has users creating a large file, to which fio performs a lot of I/O within that file. By design and by definition, that is COMPLETELY different than say creating 10,000 files, in 500 different directory of different sizes, and then opening, closing, updating, deleting and creating new files. The type of testing the fio recreates, is similar to how a database like SQL Server, Oracle, MySQL etc. would utilize a "file". That is perfect if you are attempting to understand how a database would perform I/O on a filesystem. This is NOT AT ALL appropriate if you want to understand how real file system usage would perform.