I just discovered how important the size of the clusters can be in a file system.
I backuped my macbook’s hard drive today to an external hard drive formated in FAT32. So far, nothing special. I format the computer, reinstall Leopard with an internal hard drive in HFS+ so. Then I copy back the data from the external disk to the macbook’s disk, and, oh surprise, the size of the two folders, the source and the destination, are barely different: 62,29Go on the HSF+ drive, and 86,63Go on the FAT32 drive.
OH MY GOD. Error while copying? I delete and try again (it takes 2 hours for each copy, but I was not in a hurry). And same result. I perform some research on the internet, google, ect… And here is the problem: the size of the clusters on a FAT32 file system is 32kB, and it’s 4kB on a HFS+.
A cluster? It’s the unit of storage on your file system. That means that the space used by a file in your drive will be a multiple of your clusters’ size. And that’s the problem. Let’s take a little 1Ko file, it will hold actually 32Ko on a drive formated in FAT32. And the 31 other kilo-octets? There are lost (it’s called slack space).
Conclusion: the smaller the clusters are, the less space you loose. BUT (there is a BUT), small clusters involve a higher rate of fragmentation and increase time to read and write data.
Personally, I’ve never noticed the difference in access time (even if there is one, it seems insignificant). But a difference of almost 25Go used space in my case (so a loss of space of almost 40%!!) that is not insignificant.