Your filesystem of choice?

We use XFS at work (and I work for a major London-based film and television post-production/visual effects facility) as it's an excellent journalling system which was originally introduced to us through SGI equipment (which has now gradually been phased out of the production side - and is only really used for the Infernos). XFS is pretty tough and we've not had many problems using it.

Regards,

Martyn

Dani AI

Generated

This thread is from 2004, so the landscape has changed. Modern Linux choices that matter today are not just ext2/ext3/XFS/Reiser — ext4 became the mainstream successor to ext3, btrfs brought copy‑on‑write snapshots and checksums, and ZFS (OpenZFS) remains the go‑to when end‑to‑end checksumming, pools and mature snapshots are required. (kernel.org)

Pick by workload, not by lore. For large, streaming media stores or high‑throughput archives XFS is a common, well‑optimized choice (it scales well for big files and parallel IO) but note XFS can only grow — it cannot be shrunk — so plan capacity accordingly. For general desktop and server use ext4 is a solid default (extents, delayed allocation, faster fsck than older ext filesystems). If you need snapshots, checksums, send/receive or built‑in volume features, choose btrfs or OpenZFS. (docs.redhat.com)

Operational tips you can apply now: use LVM or a volume manager so you’re not stuck resizing raw partitions later; schedule scrubs/sends for COW filesystems (btrfs/ZFS) and keep regular backups; reduce unnecessary atime writes with mount options (or rely on the kernel default relatime). For XFS specifically, use the xfs_fsr/xfs_scrub tools for defragmentation and online checking, and use lvextend + xfs_growfs when extending logical volumes. Example quick commands:

# extend LV then grow XFS to use new space
sudo lvextend -L +100G /dev/vg0/lv_data
sudo xfs_growfs /mnt/data

# mount with fewer atime writes
sudo mount -o noatime /dev/sdb1 /mnt/data

(Those commands are examples — test in a safe environment and back up first.) (docs.redhat.com)

A historical note: several people here recommended Reiser/Reiser4 back then; today Reiser has been deprecated/removed from many kernels and is not recommended for new deployments — prefer maintained, well‑tested filesystems. Match the filesystem to your workload, test with your I/O profile, and keep good backups. (phoronix.com)

Acknowledgements: this expands on points raised earlier by , and and focuses on practical, modern tradeoffs rather than historic preferences.

Recommended Answers

All 14 Replies

I used ext3 on my RedHat 7.3 box. Now I have a Mac which uses HFS with journaling.

I use ext3 on my RH9 box. I will be switching distros shortly and will probably check out a couple of the other ones. This is just a home box, so I don't know if I will see any differances.

I use ext3 currently. I tried ReiserFS originally, but decided to try ext3 for comparison on my second go-round (my Linux install started out as an experiment). I had read that ReiserFS was more processor-intensive and intended more for large database files than typical workstation applications; I have not really seen much difference as regards performance. I may try Reiser4 when it rolls out.

UFS2, baby!

That's the closest thing I could say. I primarily run FreeBSD 5.2 on my machine, running UFS2, which has journalling. But, on my Linux machines, I run ext3.

yup I was about to say, what happened to ufs? It is primarily used on aix, sco and solaris.

I know this thread is old, but what is JFS? Never heard of that one.

We were working on a distro of ours.
The best in my opinion is ufs2. But the linux kernel cannot write on it.
On Linux the best should be as of today Reiserfs and Reiserfs4 is good. I guess when it is out of beta it should be the best.

But then again this is all relative choices. A lot depends on the kind of apps and typical file sizes.


Regards
Amar

Reiserfs is fastest, ext3 is a little more safe, ext2 is fastest, different filesystems for different jobs, oh and UFS on solaris with logging on.
/ - ext3
/boot - ext3
/var - ext2
/home - reiserfs
/tmp - ext2
/usr - reiserfs

UFS is SO much better than all the FS's listed here, however it is only a Unix FS *shrugs*

I think that is very much personal opinion, and depends on system type, for example in my opinion UFS on solaris is much better that UFS on openbsd, since Sun have made so many alterations to the version of UFS that they use, to the best of my knowledge UFS on bsd isn't journalled.

Additionally as I seem to recall ext2 (which is most similar to ufs) benchmarks faster than UFS (although ufs can be more stable), and reiserfs benchmarks at about the same speed, please note that is based on memory since I can't find the benchmark results at the moment.

i would gracefully raise my hand for reiserfs..

the only thing i prefer it over ext3 is shutdown problems due to power failures and squid cache.

Regards

Peeyush Maurya

Well beat this NTFS, and FAT32. Ha!

I seem to recall ext2 (which is most similar to ufs) benchmarks faster than UFS (although ufs can be more stable), and reiserfs benchmarks at about the same speed...

Very generally though, keep in mind that those benchmarks vary according to the size of the files used in the tests- some filesystems will "benchmark" better than others if the testing is done with large files, while other will perform better when dealing with smaller files. This is definitely the case when you compare reiserfs to ext2/3.

Be a part of the DaniWeb community

We're a friendly, industry-focused community of developers, IT pros, digital marketers, and technology enthusiasts meeting, networking, learning, and sharing knowledge.