How many FILE SYSTEMS are out there can you name, and what are the differences between them?

[MadRat]

I've heard alot of acronyms lately for this file system and that file system, journalling and joliet, clustering and encrypting. There must be alot of them out there. I bet we could list two dozen without getting too detailed. I would just like to hear the differences, the advantages to using them, their origins, and anthing else trivial that would be neat to know.

Thanks in advance.

 

[n0cmonkey]

NTFS -journaling, pretty much Windows only. OpenBSD & Linux can also read (and maybe write, don't know) NTFS partitions.
FAT* (12/16/32 -but should all count as one ) -crap, don't use
HFS (HFS+) -journalling has been added. Mac filesystem.
ext[23] -ext2 is not journalling, ext3 is journalling. ext3 is backwards compatible. For Linux.
XFS -SGI's enterprise journalling filesystem for IRIX. Also ported to Linux.
JFS -IBM's AIX, OS/2(?), and Linux journalling filesystem.
Reiser -Linux filesystem that gets rewritten often.
UFS -Unix File System. Solaris still calls their filesystem UFS, although I'm sure it has been changed, for example they added journalling in Solaris.
FFS (OpenBSD, NetBSD, and FreeBSD versions, which I believe are technically different) -Fast(?) File System. No journalling, but instead uses the technically superior () soft updates. Generally Linux will support A5 (FreeBSD), but have poor support for A6 (OpenBSD).
UFS2 -The next FFS. Has soft updates, background fsck, and probably other improvements.
vxfs -Veritas file system. Used on HP-UX, Solaris, and maybe AIX.

http://www.csie.ntu.edu.tw/~b7506051/prjs/pmbr/getbsec/pmbr.os

The initial list there was off the top of my head

 

[drag]

ReiserFS version 3 is what is currently reiserfs in linux. I think reiserfs v4 is almost complete and it's special thing is that it's a "fully atomic" filesystem.

I guess that means any sort of file handling is done completely or not done at all, or something like that.

It's suppose to help out a lot in situations like file system crashes, it will make sure that no half-written or half-copied file is ever left on the file system. A journalling system keeps track of changes (thru keeping recordings of meta-data about the files) so that if a situation like this happens the OS can handle it in a intellegent way (fix, save, or delete corrupted files, repair filing systems quickly etc etc), and I am guessing the point of a fully atomic file system is to make sure that this sort of thing never happens in the first place.

here is some technical documentation.

 

[MadRat]

What is fsck? I see that term and never have taken the time to look it up.

 

[rbV5]

HPFS The H-igh P-erformance F-ile S-ystem for OS/2

 

[BFG10K]

FAT* (12/16/32 -but should all count as one ) -crap, don't use

FAT16 is good for very small partitions and volumes because it doesn't take up much overhead space.

 

[drag]

Originally posted by: MadRat
What is fsck? I see that term and never have taken the time to look it up.

FSCK = file system checker or something like that.

You use it to check the filesystem, usually at boot up. Some like xfs don't need it, so they have a dummy fsck.xfs so that it doesn't break start up scripts.

It's basicly the unix version of checkdisk. (or more likely checkdisk is the dos version of fsck)

 

[ArmchairAthlete]

I'm using FAT32! Yay?

 

[GhettoFob]

What does journalling do?

 

[n0cmonkey]

Originally posted by: BFG10K

FAT* (12/16/32 -but should all count as one ) -crap, don't use

FAT16 is good for very small partitions and volumes because it doesn't take up much overhead space.

Who still uses small hard drives on machines that have better options?

 

[n0cmonkey]

Originally posted by: GhettoFob
What does journalling do?

I think it tries to keep track of what gets written to the hard disk so the filesystem integrity isn't compromised during a crash.

 

[drag]

Originally posted by: GhettoFob
What does journalling do?

Keeps tracks of changes to the filing system. If you were in the proccess of transfering a file and your computer crapped out then when you boot the OS checks the journal and knows that that file was left in a bad state and hopefully can do something to fix it or recover it.

From a user's perspective it means that if you shutdown incorrectly your computer doesn't have to scan each and every file to make sure that it's ok, it just checks questionable ones, so bootup times are much quicker. Also your filing system is self-correcting and much more stable. (depends of course on how well the file system is designed of course)

That's why stuff like fat32 sucks nowadays. First off it was never that good to begin with, and secondly it lacks the most basic stability and data protection features that most other modern filing system uses.

Then you have advances in algorithms and stuff like that to deal with moving data. Currently the 2 probably best FS's around are JFS from IBM and XFS from SGI.

XFS has no need to run extra checks on boot up, it's self correcting/repairing on the fly. It's designed for media file handling so it's able to handle massive (2-3GB+) file sizes with ease.

Under Linux 64bit on supporting hardware it can have maximum file sizes of up to 2 terabytes. The maximum partition sizes can span 500 file system volumes of 2 terabytes each.

It's theoretical limit is 9 million terabytes file size with 18 million terabyte limit for file systems with a largest currently possible pratical limit of 8000 terabytes under the IRIX unix operating system.

And with the newer advances in speed optimizations is faster then FAT16/32.

I think that Suse uses XFS by default, Gentoo use to recommend ReiserFS. JFS is now free software thanks to IBM, which is part of the SCO lawsuite stuff (although JFS is obviously purely IBM technology)

 

[JavaMomma]

Don't forget WinFS for the next version of Windows, Longhorn...does it count as a FS?
Or is it just the next version of SQL Server ontop of NTFS?

 

[drag]

Originally posted by: JavaMomma
Don't forget WinFS for the next version of Windows, Longhorn...does it count as a FS?
Or is it just the next version of SQL Server ontop of NTFS?

It's probably will be different enough to be a entirely new FS. Or at least a new version of ntfs, there are a few different ntfs versions.

But then again it's mostly vaporware for then next year or two or so. Maybe with the gobs of money MS is dumping at it's reasearch divisions they will figure something out that is actually new and interesting.

 

[n0cmonkey]

Originally posted by: JavaMomma
Don't forget WinFS for the next version of Windows, Longhorn...does it count as a FS?
Or is it just the next version of SQL Server ontop of NTFS?

I haven't kept up with it. I heard it was a new FS for a while, then just a layer on top of NTFS, and who knows what next. I think we may have to wait for something substantial to come from Microsoft on this (or it has already come out and I didn't see it because I don't pay attention ).

Here is a comparison of journaling and soft updates, if anyone is interested.

 

[MadRat]

I think the main premise of WinFS is Microsoft's attempt to do a hybrid virtual file system at both the user and kernel levels, and something that does directory and object sorting by metafile tags. Most likely its going to be catchup to what the open source community has been brainstorming now for a couple decades but could only implement it with today's technology because of the bold visions of the developers. I doubt MS will invent anything new - do they ever really invent anything?