* Read them once and measure the durations for the read operations
* Read every file 10x and measure the durations for the read operations
* Repeat the read part of the exercise after a few hours.
Discuss your results.
Create a directory `/dev/shm/$USER` (replace `$USER` with your user ID) and repeat the exercise in there - you will probably need about 100 GB RAM to do this (or simply compute node) - no delayed repeat needed here (why?).