Reiser4 Is Now Available For The Linux 3.12 Kernel

[Grogan]

Never lost a single byte with ext4,,,

Heheh... I always said that about ext2. It's true. FSCK fixing errors on improper dismounting/incomplete writes looked ugly, but it always turned out OK for me. I still have one old system using ext2, simply because of all the Linux filesystems it's got the least overhead. It's just an old Pentium III, 500 MHz computer that I've been using as a fileserver and backup dialup (good old fashioned ISA modem) NAT gateway in case my Internet is ever down, for the past dozen years and it's still chugging away. (Though I did have to add a PCI SATA controller card for more modern storage capacity) I noticed that even network file transfers were slower with ext3 on that machine. (Haven't tried ext4... I quit upgrading kernels on that system at Linux 2.6.20 but I would expect it to use more CPU than ext2)

I remember when ReiserFS first came out, it was the first journaling filesystem ready enough for distros to use. I think it was Mandrake 8 or so where I first tried it. (I went back to ext2 when it blew up in my face)

I tried Reiser4 once, and it had me chasing my tail. I was using it for a junk partition for a while, where I stored music and tarballs and compiled stuff etc. I was having trouble getting some things to compile, like Glibc for example and it turned out that it was because of Reiser4. It compiled just fine on an ext3 partition. Some things would compile, but the timestamps were subtly broken and make was recompiling a lot of dependent objects. At first, in both cases, I thought my toolchain was somehow broken, but it was Reiser4. I'll admit that I never went back to it again, as I saw no benefit to using it. (Just a more complex filesystem with iffy repair tools, at least at the time)

I think ext4 is pretty good (for performance and reliability), I use it everywhere else.

[borsook]

I've tried both XFS and JFS and ran into kernel panics eventually. No idea why though. I tried btrfs and it was noticeably slower than the ext4/ext3 and I didn't notice much of an improvement otherwise, but it was a while since I've tested it. ReiserFS I didn't notice a huge improvement especially if I was trying to copy files from ext4 to it. I've heard that it has corruption issues over time. Reiser4 I haven't been able to properly try which is a bummer, because for an older file system that's had very little development in comparison, it's holding its own against the greats.

BTRFS's benefits are not really its performance but additional features, like snapshots, or volume compression. Speed comparisons are a bit unfair as EXT4 offers so much less features.

[a7v-user]

[IMG]http://phoronix.com/forums/images/misc/quote_icon.png[/IMG] Originally Posted by [B]AJSB[/B]  [URL="http://phoronix.com/forums/showthread.php?p=382457#post382457"][IMG]http://phoronix.com/forums/images/buttons/viewpost-right.png[/IMG][/URL]
I used Ext4 till there were some issues in some versions of it and i myself lost some data, gone to XFS and never look back.

Never lost a single byte with ext4, perhaps you've disabled barriers, or changed default data mode to writeback?
As for performance... http://www.phoronix.com/scan.php?pag...esystems&num=2

Actually there was a time of madness when Ext4 had barriers disabled by default because barriers carried a performance penalty.
There were a few hot threads here on Phoronix (and other places) about it at the time.
I remember it because my '/etc/fstab' had "barriers=1" in them mere minutes after reading about it.

I don't think Xfs had barrier support back then but it does now thankfully.
Oh... and Ext3 still have barriers disabled by default.
Finding some 0-length files instead of a 70+ hour save game was no fun at all.
Now I'm using Reiser3 precisely because it might not have the best performance but at least it doesn't destroy data if I get a kernel panic or accidently hit the power switch without first powering down. It's fast enough for me and since it's already in the kernel it's pretty reliable.

[profoundWHALE]

I've tried both XFS and JFS and ran into kernel panics eventually. No idea why though. I tried btrfs and it was noticeably slower than the ext4/ext3 and I didn't notice much of an improvement otherwise, but it was a while since I've tested it. ReiserFS I didn't notice a huge improvement especially if I was trying to copy files from ext4 to it. I've heard that it has corruption issues over time. Reiser4 I haven't been able to properly try which is a bummer, because for an older file system that's had very little development in comparison, it's holding its own against the greats.

[AJSB]

Never lost a single byte with ext4, perhaps you've disabled barriers, or changed default data mode to writeback?

As for performance... http://www.phoronix.com/scan.php?pag...esystems&num=2

Nope, i was using stock settings with Slackware and Linux Mint...lost data with both...and IIRC , also with Xubuntu.

Befiore i heard that was some issues with Ext4 (at that time) i even imagined that was a problem with the HDD, but it wasn't like tests proved and i still have that HDD fully operational, used both in Window& and now again in Linux Mint 15 using XFS and never again lost data.

My comment about performance was precisely because tests usually show faster performance of Ext4 vs XFS but *my* perception is the opposite in some tasks (you can actually notice in your own link that in some tests XFS is faster than Ext4) because i value more precisely the points where XFS is faster than Ext4.

I never said XFS is *overall* faster than Ext4, it isn't....however, *for me* , in special paired with XFCE, it *feels* faster.

[Rexilion]

From what I know is that Reiserfs was rejected because:

- It had a plugin system allowing proprietary plugins -> kernel devs supporting proprietary hardware
- Developers had no way to convince kernel devs why this filesystem is an extension to the other filesystems (i.e. it serves specific cases (better) where other fs's don't)

(Someone suggested to rename it to killerfs to make it more applicable for mainline inclusion.)

[diriel]

There were numerous technical reasons for Reiser 4 not going into the linux kernel. Hans butted heads with Linus and told Linus to change his kernel to suit the way Reiser 4 did things. Linus ended up telling Hans to .... /shrug

Not saying Reiser 4 was bad... but it was not being "rolled up" in a fashion that was a good technical fit from Linus point of view. When Linus tells you to change something...he means it. You will change it, or it will not go in. This is what happened.

[halo9en]

I uaed Ext4 till there were some issues in some versions of it and i myself lost some data, gone to XFS and never look back.

Never lost a single byte with ext4, perhaps you've disabled barriers, or changed default data mode to writeback?

As for performance... http://www.phoronix.com/scan.php?pag...esystems&num=2

[AJSB]

I'm also a happy XFS user even though Reiser4's tree balancing is technically and philosophically the best way to balance a tree.

I'm quite interested in F2FS's future since it tries to account for mechanical sympathy of NAND flash chips and it is designed and backed by Samsung.

Ah yes, i only mentioned XFS and didn't talk about F2FS because myself don't Flash drives but all that i heard about F2FS is praise for it so, yes, if i used Flash, i would use that one.

But with HDDs, it's XFS or bust !

[rohcQaH]

It's not about the name. reiser3 is still in the kernel and has not been renamed.

reiser4 wasn't included for a variety of reasons. It was proposed and rejected and improved and proposed again and rejected again long before the murder.

After Reiser's conviction, his company Namesys went out of business. Since then, there hasn't been much progress about the file system. The few who spend time on it barely manage the porting to new kernels, but lack the manpower for substantial improvements.

So why would they merge a filesystem that's alpha- or maybe beta-quality, when there's no sign of improvement?

To be accepted by now, it'd need at least two things:
a) someone willing to invest the time to fix all the technical issues that caused the rejection
b) someone willing to maintain it

That won't be namesys, and none of the major distributions seem interested. Why would they, given the current alternatives?