windows filesystems vs. linux filesystems
 

Hi all,

Does anyone have any benchmarks, or any study to show how linux filesystems compare with windows filesytems.

What are journaling filesystems, and how are they different from common filesystems. (XFS, Reiserfs etc)

XFS, ReiserFS, EXT3, NTFS, JFS are journalized. My gathering is that your filesystem is living in realtime, instead of "after" you stop using it for the day. My real life examples of journalized vs not: ReiserFS-->I don't have to reinstall the OS after a power failure (I live in the country). NTFS-->generally the same as ReiserFS, FAT32-->reinstall ater a good outage, especially with a few false starts, EXT2---reinstall!

A journallized FS is also usually faster, as the files are catalogued in the journal. I'm a fan.

NTFS is one of the all in 1 FS's (or trys to be), not a very good thing to do

i prefer reiserfs....

and theres not to many windows FS's. most of them were just dos FS's that were slightly modified, i think NTFS was the first FS they actually might have made

I dont really understand what you mean. I think every file system "lives in realtime"

I really don't know what I meant either.

With journelling. the file system is ALWAYS in a stable state.
you can cut the power at any time, and although files being writen to or modified maybecome corrupt, the filesystem will still be perfectly intact.

with non journalling file systems, dureing some stages the filesystem is not intact, but is then quickly fixed. if the power fails at a bad time, you will need to re-format.

i like reiser, if you pull the plug all that happens is the last thing you did to the hard drive is undone at worst..

"Does anyone have any benchmarks, or any study to show how linux filesystems compare with windows filesytems.

What are journaling filesystems, and how are they different from common filesystems. (XFS, Reiserfs etc)"

Well first off, most "modern filesystems" (XFS, Reiserfs, and many more) are journaling. This, for most practical purposes, is actually a dividing line. NTFS kind of falls in the middle, meaning you can crash it and it will probably work, but I wouldn't intentionally do so, or trust my data to NTFS's abilities. NTFS itself is fine, it's the other items in windows that sometimes prevent NTFS from doing what it's meant to, for "performance" reasons.

On a more basic level, a normal filesystem will wait till the most opportune time to flush data to disk. This is a performance increase. An older filesystem (ext2, fat, vfat, etc.) will simply drop everything if the power fails. "Journaling" means dropping data to an optimized part of the drive, then pushing the data to the proper spots when appropriate. Instead of finding, then opening/altering each file on disk, the journal simply records the changes as needed, preventing corruption if the computer is shut off. If the power is shut off while the computer is updating the journal, you might lose the journal, but the rest of your system is fine. If your system is updating a file from the journal when power is lost, it has the journal to help it "guess" which file was corrupted, greatly speeding repairs.

None of this means you'll regain your lost data, but it does mean you avoid several minutes of drive checks during boot. In the old system, skipping these checks was possible too, but at the risk of much bigger problems down the line, because when the FS hit a problem, that error sometimes propagated through the drive.

For the record, many filesystems are comptemplating database like systems, which has the advantage that a corrupted system might be able to extrapolate missing data using certain algorithms. This is great in theory, but it also bears the risk that a relatively simple, but severe, corruption can wipe out your entire drive. The problem's being worked on by several groups, but MS is making vaporware threats to include this with longhorn...another reason to avoid that software, as if there aren't enough already. :)

In an interview with Theo de Raadt, the founder of OpenBSD, he discussed the filesystem issue. He referred the interviewer to this paper about journalled filesystems versus the *BSD softupdates. He didn't seem to think that ext2 was a very good idea at all. Its a good read.