1

I am defining the criteria to create an optimal folder structure for cache files on the filesystem.
The aim is to create the most performing file hierarchy from a filesystem perspective.

Files cached are mainly HTML pages, so they are small files, but as far as now my searches worked out that it is not the size of files to stress the index of a filesystem table, it is the number of files inside a directory.

Identifying the ideal number in 200 files per directory, I thought to create 10 subdirectories and 180 cache files, replicating this pattern in any subdirectory, but I admit it is a purely random attempt, not driven by a real knowledge of a good methodology to calculate it.

Can anyone suggest me which are the characteristics to evaluate in this decision, or share an authoritative resource speaking at an academic level about filesystem optimization, possibly applied to great file trees?

yodabar
  • 4,651
  • 2
  • 33
  • 38
  • 1
    The hashfs (http://storageconference.org/2010/Papers/SNAPI/4.Lensing.pdf) paper from SNAPI'2010 might interest you. There we looked at a file system for such an scenario. My guess is that is is not so much the number of files in a directory, but the amount of non-cached file system metadata (dentry pages, inodes). – dmeister Mar 04 '13 at 16:11

0 Answers0