Extra features

Weigh up a range of other capabilities.

Although we undoubtedly value speed over pretty much everything else when looking at the merits of any particular filesystem, sometimes that isn’t enough to differentiate any one over the rest. In that situation, we need to take other things into consideration. For instance, you might be choosing a filesystem for a more complex setup, such as a RAID array or an LVM volume, in which case you need to investigate each filesystem’s more advanced features. On top of this advanced extras such as automatic checksums and snapshot options all help secure your files and provide additional security. Let’s take a closer look at our five contenders to find out whether they have anything else to offer.

Reiser5 8/10

About a year ago, Edward Shishkin – the sole Reiser5 developer – reported that he had stabilised the filesystem’s logical volume functionality. So, the real power of Reiser5 doesn’t lie in formatting a single plain drive partition. Instead, it’s found in enterprise-grade usage, where several storage devices are united in one logical device. The logical volume manager in Reiser5 supports an impressive feature known as Data Tiering, which involves dumping peaks of I/O load to a proxy device. The idea is to add a speedy drive to an existing array of slow drives and make the whole array perform quickly. As long as I/O peaks don’t last for too long, the Reiser5 volume balancer has enough time to flush the cache and make the fast device available for the next peak. The ability to balance data between parts of the volume and fast data migration make Reiser5 very promising.

Ext4 3/10

There’s no urgent need to tweak your existing Ext4 mounting options because your system is already using the best defaults. So, in this regard, there’s not much to explore here. Ext4 is rock solid, very fast (although it tends to lose out to Btrfs for first place) and reliable. Ext4 rarely fails, and when it does, chances are that the real source of the problem is the drive itself. Ext4 also benefits from a lower fragmentation rate than most other filesystems, thanks to delayed block allocation. That means that, even in the long run, an Ext4 partition won’t show signs of slowdown, even in the case of intensive reading and writing. However, as this part of our Roundup is dedicated to extra features, which Ext4 lacks, we’ve given it a low score here. When building an advanced storage solution, you might want to consider other filesystems, which may turn out to be more beneficial.

Btrfs 10/10

Btrfs has an abundance of extra features that put it in the top spot. First of all, there are snapshots – readable and writable copies of the entire Btrfs subvolumes. You can use such snapshots as backup copies of your system, which is a really useful feature. Snapshots can’t help in the case of drive failure, but they do prove useful if you need to roll your system back after a bad update, or in order to restore a missing file. Unless you create a large Btrfs pool and rotate its snapshots too frequently, your performance won’t degrade. However, Btrfs does hog your CPU more than other filesystems, and it also devours the effective free space on the drive. In return, it gives you transparent compression (two copies of the same file occupy the space of one), plus a few more advanced features, such as its built-in multidisk RAID support. For this it’s in welldeserved first place.

XFS 5/10

XFS has its roots in server storage, where it was used in large arrays. Before 2011, it also supported larger drive configurations than Ext4: 500TB versus 16TB. However, these days that limitation is no longer valid as the resizefs program that serves Ext4 partitions uses the 64-bit flag by default, enabling it to span up to 1,024 PiB. XFS supports online defragmentation and resizing of the mounted filesystem. It also claims to have sophisticated metadata read-ahead algorithms, extensive run-time metadata consistency checking, and robust repair utilities. From 2014, XFS applied the newer on-disk format (v5), which includes a metadata checksum scheme called Self-Describing Metadata, making it more better reliable. That’s great for production environments but not overly impressive. Therefore, there’s no explicit advantage over Ext4 and no extra features compared to Btrfs or Reiser5.

NTFS 2/10

While the NTFS-3G code seems to be in good shape, the overall performance of this filesystem in Linux is abysmal, and our tests show it. The FUSE-based NTFS-3G driver can’t win any performance test, regardless of the type of load. It looks as though it has no chance here either, as NTFS-3G has effectively no advanced features. The driver has been advertised as a high-quality solution for reading and writing data on NTFS partitions. Indeed, it doesn’t corrupt files and if a Windows partition has left NTFS in an unsafe state (for example, Windows wasn’t resumed after suspension), the driver refuses to mount it. You can happily use the current NTFS driver in Linux to check existing Windows NTFS partitions and safely write data to them, but that’s about it. Paragon’s forthcoming in-kernel NTFS3 driver should be much faster, but we don’t expect any extra features there, either.