I have an application where a large number of files have to be stored on a server for reference purposes. These documents are named from 00000000 to 99999999. Currently there are only 2 million documents, but the count is increasing.

Storing all of these in one single directory seems to be relatively slow and occasionally I have noticed some entries simply disappearing. I always have a backup so putting them back in place is not a problem, but I don't like to have them disappear.

I am wondering if anyone knows what the best way to store these documents in the file system is. I need to have a balance between quick access, stability (nothing spontaneously disappearing), and ease of backup.

In a situation like this is it more efficient to create a directory hierarchy of say:

/1/2/3/4/5/12345678 for storing document number 12345678? If this is efficient is there a certain hierarchy point to stop for efficiency purposes? For instance is:
/1/2/3/4/5/12345678 better, worse, or no different from
/1/2/3/5/5/6/12345678

Thanks for any input.

In order to balance the number of files per directory it would be better to number the directories by the least significant digit. For example file 12345678 should be in directory /8/7/6.

------------------
Steve Stites

for that many i would split into say 100 dirs based on

12345678%100

(ie the last 2 digits)
so that one would be
/78/12345678

or 1000 dirs

/678/12345678

i cant see how many levels woudl be more efficiant,
all it would do is eat up your inodes

I'd go
/123456/12345678
or
/12345/123454678

it's less confusing to people if you use the leading nums as the dir

that is true,
but that way provided that they are generated in sequential order the lower directorys fill up first.

of you sort by the last numbers then the directory load is ballanced