0

I'm formatting an external hard drive with gparted.

The original NTFS read 232.28 MB used.

Now, with Ext4, it reads 1.92 GB used.

Questions:

  1. Why?
  2. Is there a better file system I should use for external drives?

(This drive will only be used with linux computers. )

Thank you!

YQ002lc2
  • 45

1 Answers1

2

I assume those "used" values are both for a freshly-created, empty filesystem of the same size, and that's where your confusion comes from. Those "used" values suggest that the actual size of the filesystem on the external drive is probably quite large (say, more than 1 TB?).

On a freshly-created empty filesystem, the non-zero "used" value indicates the disk space allocated for filesystem metadata.

NTFS seems to store most of its metadata as special hidden files. That probably allows the space allocated to metadata to easily grow as needed, and so the filesystem does not need to allocate all of it at filesystem creation time.

On the other hand, ext4 is fundamentally based on a fairly classic filesystem design, where e.g. all the inodes are pre-allocated and the ratio of inodes-per-megabyte is set at filesystem creation time and cannot easily be changed afterwards. As a result, all the metadata space the filesystem will ever need (while at its current size) will be allocated as part of the filesystem creation process.

Using NTFS on an external drive that will only be used with Linux systems does not really make sense: on older Linux versions you might be forced to use NTFS-3g, which is a FUSE (Filesystem in User-Space) driver, and ... not exactly a performance-optimized solution. On very old or hardened systems, there is no guarantee that a NTFS filesystem driver would be available at all.

On the other hand, ext4 is very well supported and has excellent compatibility features. If you know you'll need to work with very old Linux versions, you could disable some of the newer filesystem options or even create the filesystem as ext3 or ext2 to allow even extremely old Linux systems to fully access it.

And since the ext2/3/4 has always been a kernel-based driver, it can achieve good performance (with certain caveats though: if you disable the dir_index filesystem option to achieve compatibility with old systems, directories with very large numbers of files will be slow).

However, you should note that using any Unix-style filesystem (like ext4 or XFS), the UID and GID numbers used on the system that writes files to the disk will be preserved in the file metadata. If you need to ensure all the files on the external drive will be easily accessible on any Linux system the external drive might be plugged into, see this answer I wrote in 2018.

telcoM
  • 96,466
  • Thank you for educating me! Yes, it's a large drive and yes I was confused about the difference in the used size for an empty drive. That makes sense though. Most of these linux systems use ext4, so they can handle it. For the second part, I didn't know if there were any advantages to using another file system for an external drive with linux systems or is it more trouble than it's worth? (like zfs, btrfs, etc.) – YQ002lc2 Aug 22 '23 at 22:47
  • 1
    ZFS or btrfs might be unnecessary complications for a simple external drive, and might not be available on all Linux systems... unless you already use either of those on your systems and want to e.g. transfer ZFS datasets from one system to another. So it depends very much on what you're going to use it for and the age of the systems you use it with. If you want something relatively simple but more modern (=more efficient for multi-TB filesystems) than ext4, then XFS might be a good choice. – telcoM Aug 22 '23 at 23:46
  • Ok, thank you! I'll have to play around with XFS a little bit. – YQ002lc2 Aug 23 '23 at 06:22
  • since the ext2/3/4 has always been a kernel-based driver actually there's FUSE ext4 driver which is required in case you need to mount an ext4 partition with block size larger than page size – phuclv Aug 24 '23 at 10:37
  • @phuclv Good to know - but I think that would only be needed if you have a filesystem with exceptionally large block size or other special circumstances. Apparently the only systems that could use >4k block sizes with a kernel driver would be 64-bit PPC or ARM systems with specific kernel configurations. I would expect the kernel-based driver to win in terms of performance, so it would be smart to use it whenever possible. – telcoM Aug 24 '23 at 10:48