We applied our testing technique to the Intel 530 Series SSD. Looking at the performance charts, there’s only a subtle difference in how Ext4, Btrfs and XFS perform. You could pick any of those and get the same performance; they’re all very good. If you need a reason to choose one filesystem over another, read and write speeds won’t suffice, so consider extra features (see overleaf). Reiser5 showed a lacklustre performance during the read test, though it was still decent – at least, it was better than the old Reiser4 figures. Nonetheless, Reiser5 ran close to the winners, just 10 per cent slower in the dataset test, and about 30 per cent slower in Postmark.
NTFS showed very good results in Sysbench but otherwise it was poor and couldn’t perform anywhere close to its competitors. In Postmark it was six times slower for reads, and 2.5 times slower for writes. We first thought that the NTFS-3G driver from Tuxera was to blame. It’s known to be a user-space driver that works with the FUSE subsystem in Linux and not normally about speed. Still, that’s no excuse because most Linux users don’t have anything else to choose from and are forced to deal with NTFS-3G. Paragon’s much faster NTFS driver will soon land in the Linux kernel, but until then we’re stuck with that slowness.