Quote:
Originally Posted by jefro
There have been some efforts to make a filesystem that targets a ssd but I don't know how the metrics on that stack up today.
|
Well that's one of the things that is sort-of easy. Nilfs2 and F2FS are intended for SSDs and BTRFS has an 'SSD Mode' (although it doesn't seem to do much in straightforward perf testing; more difficult to say whether it does anything for write amplification).
So something like this (
EXT4 / F2FS / Btrfs / XFS On Early Linux 4.10 Kernel...the title is slightly deceptive though as it actually tests briefly several kernel versions; there are earlier tests on different versions if that's what you want) set of tests has a series of results for various filesystems.
More relevant to the OP's system would be something like
this testing which is at least performed on a hard disk.
One thing that is at least worth pointing out is that the mount options obviously make a performance difference and as a general rule the more advanced the filesystem the more scope there is for playing with mount options. The phoronix testing is generally performed in the default condition but
here is a comparison for BTRFS.
Note that in certain specific cases there can seemingly be a big win but not all of the tested cases will be robust against other problems. Also note that compression of any kind is likely to work better in performance terms when you have 'lots of CPU/slow disk' and less well for 'fast disk/constrained CPU availability'.
@masinick
Quote:
I believe that SUSE is a distribution that goes against common convention - years ago I believe they used ReiserFS when most others were using ext2 or ext3. Recent versions of openSUSE have successfully used btrfs.
|
For about the past three years openSUSE has recommended BTRFS for the system and XFS for user data (/home). They have the ability to roll back OS updates if necessary. Given some of the comments I see on the 'net about XFS being more delicate than ext4 I don't think I'll be hurrying to try that out but it could all be FUD depending on how the XFS volume was mounted (which is not usually info that you get from people complaining of data loss).