Read "Rethinking File Mapping for Persistent Memory"
and answer the questions below.

1. Imagine implementing "HashFS" on a disk (by simply replacing the PMem with a disk).
   Will this "HashFS-disk" work well?
   If yes, why doesn't Unix do this?
   If not, what can go wrong?


2. Why both per-file mappings (extent tree and radix tree) are tree-based?
   Why not a hash-based per-file mapping? What can go wrong?


3. (similar to the above question)
   Why are both global mappings based on hash tables?
   What can go wrong for a "global radix tree"?


4. Write down the pros and cons of the four mapping designs:
   a) per-file extent tree
   b) per-file radix tree
   c) global cuckoo hash table
   d) HashFS