Linux 5.0 File-System Benchmarks: Btrfs vs. EXT4 vs. F2FS vs. XFS

[patrakov]

But... reflink with XFS is stable or not?

It is, since Linux-4.16.

[F.Ultra]

And what happens when KeePassX fills in a password in Firefox and that gets swapped out to disk? Or if it crashes and ends up in a coredump?

Well now my systems are configured to not produce a coredump but who exactly are you protecting your precious data from, Spectre or is it Hydra?

Previously this was "what happens if a thief steals your computer" and now it's "what if mega clever hackers are actively trying to steal that one password that they just hope will be flushed to swap". What will be next, implying that I'm crazy enough to use a laptop?

[F.Ultra]

i hope all the full-disk-encryption guys are properly secure their running services which are much more in danger, but given how many anti-systemd trolls not realizing namespaces and cgroups not available their holy 30 years old initscripts are missing the most likely attack vectors are wide open

Not to mention how horrible insecure the entire PC architecture is hardware wise, there is just no limit to how or where you can plug in hardware key-loggers and what not.

But then there are guys that is going completely of rails when it comes to security (and since they do for good reason I don't really blame them), e.g I have a friend that is the CEO of a VPN service where his programmers takes security quite seriously: E.g they all use laptops running Qubes and when they bought them they filled the tip of every screw with translucent nail polish and mixed in glitter which creates a unique pattern with extremely high entropy. Then they took zoomed 4K images of each screw and stores them off site and once in a while they bring back the laptops there and manually confirm that the pattern is identical.

Now that is commitment

[F.Ultra]

and then proud because uptime of 900 days with no kernel and glibc updates ever got loaded :-)

Nah, I said that they took security seriously and they do. That was not an attempt of me to mock those particular people, however many that think that they take security seriously does exactly what you describe.

[Yalok]

Is that safe to apply on SSD's? My SSD is modern (this is a 2018 laptop).

It is even healthier for SSD to run with noatime, but the other options may cause data loss in case of system crash or hard reset/shutdown.
Anyway, on my home laptop I don't have anything critical and I always do backups, so it was never an issue for me for the past 5 years.

[lsatenstein]

I use a mix of xfs and ext4. I use xfs for / and ext4 for /var. My reason is based on crash recovery experience.
With ext4, crash recovery (power failure) is pretty fast, and complete.
My experience with xfs following a crash (power failure), is that it takes double the time to recover, when compared to ext4. Like the posted results, my own experience with xfs providing faster I/O than ext4 is confirmed.

Since I like fast program loads, I maintain / as xfs based. Reason-- inserts, updates and deletes for / are relatively low in number. If I discover in a future time that the elapsed recovery time for an xfs failure is about the same as it is for ext4, then I will move most of /var to xfs
So, For my next Linux 5 installation, I will do the following:
/boot ext4
/ xfs as / is low volatility
/etc xfs ditto to above
/var xfs with high activity directories installed onto a ext4 partition and links made thereto (/var/log for example).
/home ext4

I did note that an empty xfs partition trivially uses less space than does an empty ext4 partition

[Yalok]

I did note that an empty xfs partition trivially uses less space than does an empty ext4 partition

It possible to tune ext4 to use less space and match xfs.

[akuma]

Hello, according to btrfs (https://btrfs.wiki.kernel.org/index.php/FAQ) relatime in the default mount options with COW are exceptionally bad together since the atime updates are being cow'd as well. Did you run the benchmarks with the default on? If so I would be very interested to see how brtfs fairs with noatime.

[Marc.2377]

IMO, these tests are meant to track the FS performance. With dm-crypt/LUKS, you introduce too many variables, such as cypher, AES-NI accel, etc. So I like the test as is.

That said, I am curious about if any results would be different with LUKS. I'm guessing the penalty would be about linear straigth-up, and no change in ranking. But I am just guessing.

I tested myself. See answer at superuser.com: Linux LUKS and choice of filesystem

[loganj]

Crazy! This is one of those things you don't think can happen to you, until it does. What if someone steals your computer? You want them to have access to everything?! With modern CPU's encryption costs practically nothing, there is no reason not to do it.

if someone steals your computer i doubt he/she will care for what you have on your hdd.
plus there are so many of "what if" for example what if the person that stole your computer can decrypt your hdd?