XFS is very reliable. In fact, it might be more reliable than ext4, although that is debatable. XFS is also the primary file system supported by Redhat. There are a handful of Redhat employees whose job is exclusively to work on XFS. edit: All in all, the differences are minor. You won't really notice either file system and things should just work. Both file systems are very reliable and efficient. More on reddit.com

Not an answer, just some information. Also, I have practically moved entirely off EXT4 so Ill compare to btrfs. IMO (and I too, have not explicitly benchmarked all the various fs I use) they are on par. For that matter so is BTRFS. Is there a reason to use XFS? I like to think of the extended advantages/disadvantages of the operating system. This is mainly because performance differences tend to be truly negligible in "general" use cases. However - very explicit use cases can see performance differences. But Desktop, webserver, game machine, general storage (NAS type implementation) are not that. XFS should in theory handle V-v-large files better. Btrfs can generally better optimise and handle smaller files better alongside large files. flexible inode. XFS cannot : No built-in RAID, No built-in LVM support. XFS is: Journaling, Direct I/O, Parallel I/O. BTRFS is: Copy on Write, LVM, RAID, built-in filesystem compression. https://linuxhint.com/comparison-between-btrfs-and-xfs-filesystems/ XFS started around 1993, and has been updated over the years to deal with changes in the ecosystem. EXT4 was the next generation from EXT3 (and EXT2). Released in 2008. BTRFS was released in 2009 initially, declared stable in 2013. Has some serious backers (Facebook, Fujitsu, Fusion-IO, Intel, Linux Foundation, Netgear, Oracle Corporation, Red Hat, STRATO AG, and openSUSE) More on reddit.com

I've been using ext2 / 3 / 4 since 1994. It's stable and works fine. The last time I had any issues was in ext2 before ext3 added journaling. I've never bothered trying XFS but it should be fine. I tried btrfs 10 years ago and ran into a bunch of problems. I use it on some laptops now but I don't use any of its advanced features. In 32 years of using Unix systems I have never run out of inodes. You can type "df -i" and see the number of inodes you have and how many you are using. On my media drives with a relatively low number of huge files I formatted them with largefile4 so I have even less inodes than normal. I'm still only using 2% of my inodes. https://wiki.archlinux.org/title/ext4 mkfs.ext4 -T largefile /dev/device mkfs.ext4 -T largefile4 /dev/device If you don't specify these largefile/4 options and just use the defaults I would be very surprised if you run out of inodes. By default this 10TB drive should have around 233 million inodes. I use the largefile4 option to give me more space for data and even though it is 91% full and I'm only using 2% of the inodes. df /media1 Filesystem Size Used Avail Use% Mounted on /dev/sdk1 9.1T 8.3T 909G 91% /media1 df -i /media1 Filesystem Inodes IUsed IFree IUse% Mounted on /dev/sdk1 2.3M 37K 2.3M 2% /media1 More on reddit.com

It depends on a few factors like type of storage, if the app is single or multi threaded, etc. From my experience for example on basic low end ec2 storage it makes no difference performance wise, but if you use higher iops it makes a difference and xfs wins. Same for vmware esxi; a vm with the disk on fast storage has better performance with xfs versus for example a disk on a datastore over nfs where there is almost no difference. Test your workload with both and decide is my suggestion. https://access.redhat.com/articles/3129891 More on reddit.com