Best Filesystem of 2023

[@CaptainDangeax]

Always very interesting. XFS is my long term choice, let's see what's new

[@llothar68]

Is it finally doing well with small and tiny files ? This was a real killer feature for using XFS

[@CaptainDangeax]

@@llothar68 I don't deal with small tiny files since I stopped to search the marginal gain using Gentoo

[@CaptainDangeax]

@@llothar68 i don't have small and tiny files problem since I stopped to use Gentoo for marginal gain

[@CaptainDangeax]

Since I stopped Gentoo, no need for small and tiny files anymore. ReiserFS was very good in this subject though

[@pepeshopping]

No ZFS in a “modern” file system comparison? Is BTRFS finally “finished”?

[@CyberGizmo]

nope but they are working on it. pepe

[@lsatenstein]

@@CyberGizmo Stratis was supposed to be the replacement for btrfs. It seems that they are a match. Choosing Stratis means you give entire disks to the distro, whereas a btrfs volume is able to be allocated just a few gigs. In my view, btrfs vs stratis, from the benchmarks, is exchanging 4 quarters for a dollar. DJ, in this great study did not include data recoveries after a crash, I think that doing so would eliminate all but the two I mentioned, Crash recovery was not addressable by DJ because, he, in my grandfather view, does not have a lifetime to dedicate to this study.

[@Waitwhat469]

@@lsatenstein I think the btrfs -> stratis transition is more on a developer cost than the better system (by focusing on reusing existing tech and up streaming). Hopefully that transitions to a better fs as time goes on though.

[@DavidLindes]

What are the units for the x axis on all these charts (starting at 3:53) - is it a unit of time? a unit of (some multiple of) bytes per time? i/o operations per time? some sort of hybrid score? other? Gotta label those axes! You seem to be implying that bigger numbers are better, and I think I see some stuff that's in KB/s in iozone documentation, so guessing that. In future, though, a humble request that you please do include that information on your charts, because it's really hard to glean meaning from them without some sort of knowledge of what different numbers mean. :) (The FIO charts starting at 21:13 do a nice job of this, with the headers up top about MB/s, IOPS/sec, "More is Better", etc.) Still, thanks for this. I've been pondering what I want to use for some new systems, so it's nice to take a look at this.

[@CyberGizmo]

Yes iozone more is better on their benchmark tests always, and yep its MB/sec (most people today don't grok KB/sec) I tried and the labeling looked awful, but was so tired at the end forgot to mention it in the video what the units were...takes a lot of time to run those tests about 18 hours total. So any mistakes I blame on blind fatigue at that point...:P

[@DavidLindes]

@@CyberGizmo fair enough. Well, thank you for all the effort! And for replying here to fill in the gap. :)

[@Vednier]

"BTRFS, It does OK" - is most common thing i heard about it.

[@D.u.d.e.r]

Thank you for making this video. It really helped me to understand pros and cons of these file systems and best purposes they are used for.

[@katrinabryce]

zfs for me, because not losing my data is far more important to me than performance. Performance isn't that bad. Certainly for reads, I have a 95%+ hit rate on the ARC, so they are basically at RAM Disk speeds.

[@terrydaktyllus1320]

No, doing proper backups stops you losing your data - simply relying on a filesystem (or RAID) as a backup is very foolish.

[@katrinabryce]

@@terrydaktyllus1320 Backups are also important, but zfs guarantees that your backups will be of the actual data you saved to the disk, and not a corrupted version of it. It doesn't guarantee you will get your data back, but it does guarantee that you won't get wrong data back.

[@terrydaktyllus1320]

@@katrinabryce Wrong again. It depends entirely on the backup process used, and verification, to determine whether or not one or more files are corrupted. If you are relying entirely on the filesystem to provide a "single point of failure" for file corruption then you are just setting yourself up for failure. I suggest you go read up on "disaster recovery" and the "I" in the "CIA Security Triad" that stands for "Integrity" - as in, "data integrity".

[@katrinabryce]

@@terrydaktyllus1320 It provides an additional check that most other file systems don't offer. The risk it defends against is non-catastrophic failure. Catastrophic failure you know about because when you ask for the file, you get a read error. Non catastrophic failure is when it gives you something in response, but it is the wrong thing. With other filesystems, you will just back up that wrong thing. Of course you need backups too, but you can't rely solely on them either, that is just a different point of failure. zfs will let you know more often that you need to be recovering something from your backup.

[@terrydaktyllus1320]

@@katrinabryce You're not getting this are you?. "The risk it defends against is non-catastrophic failure." ext4 does that already - that's what journalling protects against and ext4 has journalling. You have reading to do. "Catastrophic failure you know about because when you ask for the file, you get a read error." Yes, ext4 does that too. Plus the backup system does it too. Next? "Non catastrophic failure is when it gives you something in response, but it is the wrong thing." Yes, it's called "file integrity" as a cybersecurity consultant for the past 15 years who deals with disaster recovery. And, no, the definition of "non-catrastrophic failure" is the ability to restore data loss from journalling in the filesystem OR a backup - as I have already stated. "With other filesystems, you will just back up that wrong thing." Not necessarily - mirrored RAID systems might have that problem but striped RAID with redundancy (which is much more common anyway) has protections against that too. You have yet more reading to do. "Of course you need backups too, but you can't rely solely on them either, that is just a different point of failure" No, I never said that. I have two potential points of failure - the filesystem or the backup. Incremental backup systems also have checksum protections to check newly backed up files against existing ones. Now you need to read up on backups too. "zfs will let you know more often that you need to be recovering something from your backup." No, it does not do that. It can recover data from snapshots that are an advanced form of journalling - but those are not backups. And, in summary, all you've demonstrated is even more reliance and far too much trust in a single point of failure - so we're back to where we started this conversation, we've gone full circle and therefore there's no point continuing further. Discussion closed, have a lovely rest of your day.

[@andarvidavohits4962]

I use single and multi-device Btrfs raid0/striped configurations on most of my everyday Linux machines and a variety of different RAIDZ (ZFS) configurations for storage but generally under FreeBSD. When used in conjunction, these two filesystems afford me just about everything I could ever wish for in a home-lab setup.

[@anonymouswizzard5680]

22:08 This spooked me. Why? Why start music with a haunting screech after most of the video was without it? That's just evil. listening at 1.5 speed i dropped a few bricks.

[@lsatenstein]

Good morning from Montreal. First of all, I want to thank you ever so much for today's presentation. I was contemplating a switch from btrfs to stratis, and now I realize that it would be the same as a 1-1 exchange, with no true benefit. I have two questions. 1) Did you work all night to produce today's top rate presentation?. I recognize that there was a lot of work involved in planning and executing it, 2) Were you using SSDs or nvme's? If so, did you also happen to run a fstrim before each set of tests? To do so would eliminate the ssd/nvme biases, however small. I am a Fedora user and I was wondering what distro you used for your testing. (not a question, but a curiosity,),

[@CyberGizmo]

Good to see you again 1. Yep, it took 2 days to finish up all the benchmarks I didn't stay up all night but worked on it from about 6am till midnight, so a long day for sure NVMe's I have all the SATA SSDs tied up over in the gluster cluster 1, and 2. yep, usually do a fstrim on the drive (i think i did it each time, won't swear to it though there were a ton of runs) Fedora Workstation 37 - kernel 6.1.7 for these runs on an Intel Core i5 12th Gen, NVMe was a Samsung 980 1TB

[@lsatenstein]

@@CyberGizmo Here in Montreal, the flu is rampant. My school teacher daughter brought it home and shared it with her husband, her three kids, and their immediate friends. Seven in one house at one blow. D.J. Please stay away from crowded rooms until March. By then the rampant flu would have dissipated in concentration, and it will again be quite safe to be in crowds. I read where Gnome 40 beta was announced. I hope it will appear with Fedora 38. Whatever topics you publish, I ever so much look forward to each.

[@osgrov]

It's quite remarkable how poorly btrfs performs sometimes, still. Makes you wonder why. At work all our database servers run XFS and storage servers are ZFS. Not sure what the reasoning behind it is (been like that since forever and I'm not an admin) but it works well. Workstations (if Linux) typically just use ext4.

[@Vednier]

ext4 is reasonable choice for general purpose workstation. Its better supported (by maintainers) then, say, JFS, its have good enough performance, its (finally) solved many long-standing problems, like slow deletion of big files (it was real pain on ext3) and so on. Its stable, yet its doesnt have many "fancy and new" features like BTRFS. However, for SSD i use F2FS, works fine.

[@gljames24]

I'm using BTRFS, but I hope to convert to bcachefs when that releases and becomes stable.

[@jlinkels]

Thanks for the effort you put in the presentation. However I miss the definition of "best". You are talking about best speed performance. Which is important. But what about other parameters? Since there are many file systems and file systems in use some should have different characteristics which make them stand out. It cannot only be speed. I am still using XFS for my USB backup disks. Millions of files. On the servers I replaced it for ETX4 because that is what Debian recommends. But actually those are the same files which are on the backups.

[@zesgod]

Where is the link of OpenBenchmarking? 21:17 Which version of Linux Kernel did you use?

[@breadmoth6443]

F2FS is flash based, so did you test all these filesystems including F2FS on a SSD or mechanical drive?

[@CyberGizmo]

Yep, one of the reasons I was complaining, I don't know if I am too impressed with the Samsung 980 NVMe SSD, so far in my testing its slower than the 970 EVO

[@DavidLindes]

I was wondering about this, too. Several things said seem to imply this was spinning rust, but, not sure!

[@deterdamel7380]

I have choosen btrfs, due to features/footprint. Some low-end devices using F2FS. But I had never any issue with a (Linux-)fllesystem in the past (ext2, reiserfs(3?), ext3, ext4, f2fs or btrfs).

[@Waitwhat469]

Some other things to add in the comparison to me would be features supported: like fscrypt support (ext4, F2FS), transparent compression support (btrfs,F2FS), and Deduplication (xfs, btfs).

[@frechjo]

The music at 22:12 sounds a lot like a remix of my friend's song: kzbin.info/www/bejne/iZuldJiiobWrkNE I had to go check if they were "reusing" his song! (Can't say for sure, I should show him this video)

[@CyberGizmo]

could be does your friend upload to artlist.io?

[@frechjo]

@@CyberGizmo not as far as i'm aware. but i listened to it again, and now i think it's just coincidence. it's the same chord progression, on a similarly sounding guitar (but played faster, and all other stuff different).

[@andreigiubleanu]

DJ is the man!

[@dezmondwhitney1208]

Good Video, D.J., This is Very Interesting.

[@CyberGizmo]

Thanks Dezmond

[@ejbully]

Baremetal tied up by the BEST FILESYSTEM... gave you the answer right there... else is second but can not compete.... ZFS is in another league of its own

[@daniilboiko]

It would be nice to hear not only about performance but about features as well.

[@paulwratt]

That weird IOWAIT errors you got on one test (the FREAD C function?), that is _not_ your drive. I see the same thing happen with the Kernel Cache/Buffer/Swap cleanup service. There is either a race condition or an erroneous timeout condition in the underlying kAPI interface.

[@bertnijhof5413]

OpenZFS right or wrong, my file/volume system since 2018 :) :)

[@merthyr1831]

btrfs hands down. Saved me today after absolutely bricking my manjaro install with almost zero user input on my part. User experience is fantastic. Otherwise no difference. Very happy.

[@terrydaktyllus1320]

It's backups that save you from losing important data, not the filesystem. A BTRFS file system with no backups is a single point of failure. An ext4 filesystem (which you may consider to be less reliable than BTRFS but I have no real problems with it in the 15 years of using it) with backups has two points of failure - therefore it's the better solution of the two. RAID and using newer filesystems are NOT backup and disaster recovery strategies.

[@merthyr1831]

@@terrydaktyllus1320 It's more that it helped save the OS when I made a breaking change rather than protecting my personal data. I agree with your points of course. I handle personal data with a Nextcloud instance on my home server with daily backups :)

[@terrydaktyllus1320]

@@merthyr1831 There's no doubt that in a lot of corporate environments that the filesystem matters when it comes to the performance of applications like databases, for example. I work on Red Hat Enterprise Linux servers on a daily basis - something in the back of my mind tells me they use XFS quite a bit but I've never really had to check. It seems to work fine. At home I've got a "plethora" of Linux systems, particularly old Thinkpad laptops, and I just use ext4 without really thinking about it. My point is less about the filesystem you use but not using its "reliability claims" as an excuse not to do backups.

[@j3kfd9j]

Does benchmarking a flash filesystem (F2FS) in a VM make sense? I assume the VM doesn't emulate the characteristics of flash hardware.... Thanks for doing this comparison

[@VishnuVardhanS]

Thank you

[@johnsmith1953x]

All this is for non-parallel FS. For Parallel FS, look at Lustre, GPFS, BeeFS (formerly FraunhofferFS) and Panasas. Which is best?

[@Waitwhat469]

Don't forget Ceph-fs on a parallel fs list.

[@rezaosman92]

Umm, which version of Linux kernel this test was performed on?

[@JohnnieWalkerGreen]

So my favorites, vFAT32 and exFAT, were not even mentioned!

[@notquitecopacetic]

I am way out of my depth here. I am just trying to learn and understand for the intrinsic value of learning. I do nothing with servers other than small homelab projects. From the outside looking in, it seems like the profiles of each system is very close, with a few exceptions. BTFS and ext4 in particular are close. Are these smallish variations enough for an admin to base their choice on, or are they looking more at the features available in each file system? Do you have a video explaining the various features of each? If you were to take an educated guess, would you say ZFS would have made the results more variable i.e. would it be an outlier in one direction or the other?

[@terrydaktyllus1320]

My advice is focus on a good backup strategy and deploy that first, then think about what filesystem works for what you want to do. Too many people (there are already a few of them commenting here) get complacent in their belief that an "advanced" filesystem or RAID negates their needs to do any backups. Whatever storage media or filesystem that you use, there is always a risk of hardware failure, corruption and data loss - RAID and the filesystem are still a single point of failure. Backups create two (or more, depending on number of backups) points of failure and therefore provide better data security, whatever the filesystem. Unfortunately, too many newbie Linux users treat it as a "fashion accessory" to impress their peers with rather than thinking about Linux in the "engineering terms" of "cause and effect" where you consider the impacts of any changes you make to a system.

[@godnyx117]

Thank you for these videos! I have seen your previous ones and they are great! I would love it if you would like to add OpenZFS and DragonFlyBSD's hammer2 in the comparison! I would argue that these two are even more valuable than XFS and F2FS!

[@CyberGizmo]

Agree completely with OpenZFS, It is my main data store and has been for about 5 years now, will look into DragonFlyBSD's hammer2 have heard good things about it for years, but never tested it.

[@godnyx117]

@@CyberGizmo Glad you are open minded! And I do admire how you are still interested about technology and even make videos! Have an amazing day!

[@johnwpierce3]

Ah the great FILE SYSTEM WARS of 2023!

[@breadmoth6443]

I use F2FS for the root partition on my SSD. I would like to see more distros include F2FS as an option since SSDs are now becoming more common place. Also it would be great if other flash-based filesystems were included such as JFFS, JFFS2 , YAFFS. Sure you can still use ext4 or XFS, but the problem with those in general is that they are written for mechanical drives in mind, where as the other listed filesystems are written from the ground up for NAND-flash drives. So it would just make logical sense imo that those filesystems should at some point be available as an option. Since I use Slackware, I am rather lucky that the maintainer has indeed at least included F2FS as an option for installation - of course the caveat is you cannot boot off of F2FS , so I had to create a smaller ext4 boot partition on my SSD.

[@Vednier]

There is actually not problem with having small boot partition witn not F2FS - its not like its getting MUCH writes, just occasional kernel update. Root is way different story.

[@CaptainDangeax]

@@breadmoth6443 after numerous tests, I got far better results for mariadb using xfs than f2fs. No problem having a distinct ext4 /boot.

[@breadmoth6443]

@@CaptainDangeax i use XFS too for my conventional disk as a home partition , i have nothing against XFS it is a goof FS too , but again , that file system was not written from the ground up for a NAND-type device. I only know of Slackware that has F2FS as an option to install on if you have an SSD , and I wish more distros would offer that, as well as JFFS2 for comparision.

[@CaptainDangeax]

@@breadmoth6443 to summarise my point, /boot ext4, / xfs or ext4 because not many writes, /home f2fs, this makes sense, /data with some database storage xfs

[@mytech6779]

SSDs are basically all cache, I'm curious how the re-read(caching) results would look on spinning disks. I know they've kind of become an intermediate archive media between the active system and cold tapes, but there are still a few live disk systems found in the wild.

[@Vednier]

That depends on that you call "live disk system". Most common case is SSD for system drive and bulk\cold data or even /home on HDD, albeit there is some transition to QLC SSD's for media storage (funny on its own, since raw QLC may be slower then modern HDD).

[@sarkybugger5009]

Looks like sticking to ext4 was a smart move. Makes you wonder why they bother trying to improve on a 15 year old FS.

[@terrydaktyllus1320]

It's because of "new broom sweeps clean" concepts - new people have to feel they make some kind of impact or change to a system, even though there is no improvement in what they come out with. I agree, ext4 "just works".

[@sarkybugger5009]

@@terrydaktyllus1320 Cough "SystemD" cough. 😁

[@terrydaktyllus1320]

@@sarkybugger5009 Indeed...

[@DavidLindes]

Well, XFS is almost 30 years old (22 for the Linux port)... By your logic, the question comes to mind of why'd they bother making ext4, then? I think perhaps this isn't the best way to think about these things.

[@mrb180]

why do they keep trying to improve a 30 year old OS?

[@tukiendorf]

Why you shown benchmarks from Stratis? Stratis isn't a filesystem. It's just an another tool to manage storage devices. Stratis is using XFS as main filesystem and LVM for disk management.

[@esra_erimez]

CephFS vs GlusterFS?

[@chamseddinegharsallaoui5495]

This how i see it : You Can get zfs under ceph or zfs under gluster so these two are more likely as container / distributed virtual disk but not a file system ...

[@esra_erimez]

@@chamseddinegharsallaoui5495 Yes, but I was trying to prompt DJ to do a video about them 🙂

[@chamseddinegharsallaoui5495]

@@esra_erimez me too....(i wanted Zfs under ceph so 2in1)

[@CyberGizmo]

Gluster for sure, Ceph and I dont speak anymore, everytime I try and get it installed on ARM it decides to quit working, works fine on Intel, but ARM....grrrrr... maybe I will try again its been awhile

[@rashie]

👍👍

[@Vednier]

18:30 - actually, this is wrong for F2FS. Its Log-based FS, so it NEVER re-writes files, instead its always new write to new place. F2FS have no benefits on rewrite, its idea is to decrease rewriting in same place as much as possible.

[@pavelperina7629]

Which is handled by controller on SSD I assume (maybe not on SD cards) ... I might try it on Raspberry, I don't even know what it uses.

[@Vednier]

@@pavelperina7629 Yes it does, BUT 1) Its never HURTS to assist controller into doing its job 2) We dont really know how wear leveling works on certain models of SSD. On SD its safe to assume there is no wear leveling at all (there is some high profile card which claim to have ssd grade NAND inside, but i didnt heard about TRIM on SD at all). I do agree that maybe F2FS isnt that critically necessary for using SSD as one may claim and yet, i used it for several years and i didnt seen ANY problem at all. So why NOT, actually? Additional good thing with F2FS is that its doesnt need periodic TRIM at all. Instead of "immidiate-after-delete" or "on demand\schedule" it monitors deletion on its onw and TRIM drive while its idle - something ext4 isnt capable of. F2FS however may be very usefull on, surprise, SMR HDDs, cause this drives strongly dislike rewriting data in same place ever. Still, this whole test is kind-of-bullshit because its virtual machine. He doesnt write on real media, so how is data actually mapped to real storage isnt known and whole point of all FS optimizations goes to window.

[@Knirin]

@@Vednier is F2FS designed for LBA mode flash versus direct access. It seems to be designed more for direct access but useable for LBA mode.

[@eznix]

I stick with ext4 because it's proven stable and the recovery tools are mature. The performance differences do not tempt me. Xfs would be a second choice. I neither want or need the features of a CoW filesystem like btrfs or zfs, so why experiment? For a single user desktop system, ext4 offers all the performance I need and I will never be able to afford a device that cannot be accommodated by ext4's feature set.

[@CaptainDangeax]

@@eznix xfs is the default fs for redhat

[@Appalling68]

Wow thanks again, DJ. Still sticking with ext4 because it's the least complicated for me, especially when trying to problem solve! Edit: GeezusHchrist looky how ext4 simply dominates in most of the benchmarks. Just wow!

[@CyberGizmo]

Yep still use it on my arm machines works great there too

[@terrydaktyllus1320]

@@CyberGizmo Yes, ext4 is the standard and very much cross-platform. If you have an ext4 formatted storage medium, there's an extremely good chance that any Linux device will be able to read it - if it's ZFS, BTRFS or some other filesystem, support for it may need to be compiled into the kernel or a module loaded.

[@tsiiphsycoii]

Rocking ext4

[@OpenSourceAnarchist]

I've been using F2FS on my SSDs for years and never have had any issues. Ext4 on mechanical drives. If I had important data, I would use btrfs, but everything I care about is backed up to the cloud so as long as Google doesn't go bankrupt or delete my account, I should be OK :) But I'm extremely excited for bcachefs and will switch to it as a test whenever it gets mainlined!

[@modjohn]

bcachefs is the most exciting in 2024, now on mainline kernel 6.7

[@fabricio4794]

DJ i wanna see your Live Set

[@CyberGizmo]

oh oops I keep forgetting to do that, need to clean it up a bit and then will do it, thanks for the reminder

[@fabricio4794]

@@CyberGizmo a DJ set..do you use Mixxx app to mix music on Linux? I use....cheers...

[@terlik3537]

Heyy I am the one watching that old video

[@RawLu.]

Please consider doing a deep dive on F2FS DJ? Am very interest to see what you think of its Stability etc.? I switched my USB's to F2FS as its suppose to be made for that? 🤔 I finally escaped decades of Nothing But Windows 👿 Everything a year or more now? and I don't want/like supporting ANYTHING Windows Based any more, if I can help it. I want to Support Linux Based 😇

[@leucome]

I use F2FS on all my SSD since 2018. Nvme, Sata, USB, It work fine I never had any issue.

[@Jeffsa12]

Why not include ZFS?

[@_sneer_]

He literally said that in the beginning of the video. Did you even watch it?

[@NeptuneSega]

@@_sneer_ small attention span folks that skip all over the place and don't allocate time to thoroughly digest the content. This is today's problem

[@DavidLindes]

@@_sneer_ There was some explanation, though I honestly don't really understand it. Like, this was all in a VM with a drive dedicated to it, so... what does it matter that it's "in use" (0:31) on the main system? And even if there is a good reason for that (that I'm just overlooking), it seems like the kind of thing one might wait for the completion of to be able to include it in a video that you have reason to believe (0:07/0:11) will be getting watched for the next year or so. So, I personally think the question is a valid one, even with the explanation. DJ, I hope you can do that follow-up (24:53) sometime!

[@_sneer_]

@@DavidLindes if the file system is in use how would that affect the benchmark?

[@DavidLindes]

@@_sneer_ "the" filesystem? Which one? I presume new filesystems were created for each benchmark, no?

[@guilherme5094]

👍!!

[@chamseddinegharsallaoui5495]

ZFS.... For me... Its number 1.... Then btrfs (as ZFS for poor) Then ext4 as Standard..... And best of is zfs under Ceph....

[@rahilarious]

Same for me except I'd choose xfs over ixt4

[@chamseddinegharsallaoui5495]

@@rahilarious as i knew ... xfs is the default in redhat 4-5 as Host for Oracle DB (10g )

[@rightwingsafetysquad9872]

ZFS for hardware, ext4 inside VMs.

[@QisHD]

Where is ZFS? This makes the whole video moot.

[@shadow7037932]

ZFS for life!

[@sotecluxan4221]

[@terrydaktyllus1320]

ext4 - "just works" and has done for years. Remember, folks, a filesystem is for life, not just for current fashions or amaze your peers at how "l33t" you are.

[@chamseddinegharsallaoui5495]

Sad but true .... The Real ZFS. From The Sun system rocks (the open one is trying , its an imitation) .... Sorry to be the Bad Guy.... And the open ZFS is doing great so far .... And i daily and many of you use it ( in virtualisation like proxmox and PBS, in NAS Synology /qnap , in Big scale storage added to Ceph)

[@seanld444]

Btrfs isn't "l33t". It has legitimate benefits, if thoss benefits are what you're looking for. Most people won't need them. But I personally enjoy having system-wide snapshots and transparent compression.

[@terrydaktyllus1320]

@@chamseddinegharsallaoui5495 Nobody is "bad" for expressing an opinion - even if it's one I don't agree with.

[@terrydaktyllus1320]

@@chamseddinegharsallaoui5495 I could also add that the reason I'm not using ZFS, BTRFS or any other "fashion accessory filesystem" is because since I started using Linux back in 1996, the ext* filesystems have never been any real problem for me. And you can't beat opinion based on real-life knowledge and experience.

[@terrydaktyllus1320]

@@seanld444 "It has legitimate benefits" and "Most people won't need them" In which case you've contradicted yourself because if most people won't need them, they are not, by definition, "benefits". "But I personally enjoy having system-wide snapshots and transparent compression." I enjoy backing up all of my files from trouble-free ext4 filesystems. I have two potential points of failure (the filesystem or the backup) but you have a single point of failure with your snapshots, whatever the filesystem. I win.

[@nilz91]

DJ is the man!