• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Best Linux filesystem for large RAID array?

J

Joe Fission

Limp Gawd
Joined
Jan 15, 2006
Messages
167
So, I'm not all that up to speed on the various Linux filesystems... it seems while I wasn't using Linux, a whole bunch of new ones showed up. I'm setting up SUSE (open source edition) v10.0 on my server and it wants to use Reiserfs for the filesystem on my RAID array... Now, not knowing exactly what the differences are between the filesystems, I'm not sure if this one will suit my purposes.

Just to give you a description of the server, it's main purpose is a file server (but will do folding in the background as well) and will basically serve files to all of the other systems on my network via Samba. The RAID array currently has a usable capacity of 896GB, but I will most likely be expanding the space as needed in the future to its max of about 2.1TB, and I want to be able to use it as one large partition, so the filesystem will have to support partitions of over 2TB. As to the size and type of files stored, there will be a lot of large video files, as well as archives and CD/DVD images, some ranging up to multi GBs in size. Other than that, I'm most likely looking for reliability over speed since the server will be serving files over GbLan, so extreme speed probably won't be required. Any suggestions? :confused:
 
XFS. Reiser doesn't really do all that well on large arrays, it takes a minute or so to mount my 980gb array but XFS mounts instantly, and performance doing large files (movies, etc) is much better on XFS.
 
XFS FTW. You'll almost definitely want a monitored UPS (one that will shut down your system nicely when power goes out); XFS does some things, from what I understand, that assume writes succeed once pushed out to disk. Caching means that this assertion can fail, potentially catastrophically. However, I've had great luck recovering from disasters on XFS (not that I'm recommending creating any :p) so I'd say as long as you've got a $50 APC thingy you'll be fine. xfs_repair is your friend, though, in time of disaster.

 
XFS hands down. It performs the best by miles when you start dealing with large files or large amounts of files. I personally think XFS is the best all-around filesystem right now...it performs well no matter what, and just gets better as the sizes get larger.

Also you won't have to worry about any size constraints...XFS uses 64-bit addressing so the maximum partition size is somewhere around 2 million terabytes. It also allows arbitrary file attributes to be added to any files which might be useful in certain scenarios.
 
I'd actually stay away from XFS and Reiserfs.

XFS doesn't handle bad sectors, loss of power, or anything of the like very well.

Reiserfs stores it's tree information in a random location on the disk. If it needs to re-build that, you have to search the entire disk.

I'd suggest checking out JFS.
 
[h]ardc[h]ris said:
I'd actually stay away from XFS and Reiserfs.

XFS doesn't handle bad sectors, loss of power, or anything of the like very well.

Reiserfs stores it's tree information in a random location on the disk. If it needs to re-build that, you have to search the entire disk.

I'd suggest checking out JFS.
Click to expand...

problem is he's dealing with both large files and large amounts of files, all on a large volume. JFS handles all of these scenarios very poorly...it bottlenecks pretty badly on large volumes with lots of files.

Although XFS isn't the most resiliant file system if he ran his server with a UPS there shouldn't be any issues. Heck, I've been running XFS on my desktop and laptop for quite some time with no issues. Though for important data I'd look into a UPS and/or some redundancy, though that is good advice for ANY filesystem not just XFS.
 
Ya, a UPS is on the list for purchase very shortly here... the main reason being that I'm using an Areca card with onboard cache memory and I don't have the Areca battery backup unit. I figure it makes much more sense to get a UPS than the BBU since the UPS will also handle minor surges and power fluctuations as well.
 
I wouldn't worry too much about bad blocks - disks these days are so dense that a few bad sectors are inevitable. They're designed to map around them, and put the data in spare sectors. And the raid controller will abstract that all away anyways - it will deal with the disks and let you know when one's being problematic. XFS will never know there's a bad block.

 
Wow I think alot of people missed the first post...

"Other than that, I'm most likely looking for reliability over speed since the server will be serving files over GbLan, so extreme speed probably won't be required."

:p

XFS is fast but that doesn't matter ;P.

For reliability I would go with ext3. No way I'd use reiserfs on a large array, reiserfs is very over rated. You know how every reiserfs tool states "this tool will destory your data" they don't lie. Its a somewhat reliable fs until it becomes corrupt and then it just seems to destory itself no matter what you do. Even as a 'fast' file system wait until it fragments itself to hell and then see how 'fast' it is.

Anyways ext3 is the most reliable file system.

XFS can be very dangerous if your system loses power. Its much less of an issue since v1.1 but it cant compare to ext3 which is journaling everything not just metadata. I am starting to like XFS and choosing to use it as a general purpose fs however its not the most reliable.

JFS is SUPPOSED to be bulletproof however I have not seen it in action. I believe there are tons of options to make it a VERY secure filesystem however unless you know what your doing I'd recommend ext3.

ext3 is the standard when comparing file systmes. Its the most trusted and the most bland. It doesn't trade anything for performance and sometimes thats a very nice feature. Anyways every file system has its advantages and disadvantages but unless you know what they are just use ext3 because you can't go wrong.
 
bexamous said:
Wow I think alot of people missed the first post...

"Other than that, I'm most likely looking for reliability over speed since the server will be serving files over GbLan, so extreme speed probably won't be required."

:p

XFS is fast but that doesn't matter ;P.

For reliability I would go with ext3. No way I'd use reiserfs on a large array, reiserfs is very over rated. You know how every reiserfs tool states "this tool will destory your data" they don't lie. Its a somewhat reliable fs until it becomes corrupt and then it just seems to destory itself no matter what you do. Even as a 'fast' file system wait until it fragments itself to hell and then see how 'fast' it is.

Anyways ext3 is the most reliable file system.

XFS can be very dangerous if your system loses power. Its much less of an issue since v1.1 but it cant compare to ext3 which is journaling everything not just metadata. I am starting to like XFS and choosing to use it as a general purpose fs however its not the most reliable.

JFS is SUPPOSED to be bulletproof however I have not seen it in action. I believe there are tons of options to make it a VERY secure filesystem however unless you know what your doing I'd recommend ext3.

ext3 is the standard when comparing file systmes. Its the most trusted and the most bland. It doesn't trade anything for performance and sometimes thats a very nice feature. Anyways every file system has its advantages and disadvantages but unless you know what they are just use ext3 because you can't go wrong.
Click to expand...

When I tried to create ext3 on my 980gb partition it supremely crapped out. XFS worked immediately. Maybe its an isolated incident or something related to the amd64 arch, but I had problems.
 
^ bexamous makes some good points as well. ext3 definitely is the failsafe choise for linux filesystems...it is incredibly safe and all-around it's a great filesystem.

However, the one things that concerns me is the sheer size of the volumes we are talking about here. I'm afraid that even routine tasks (creating filesystems, resizing filesystems, mounting and unmounting) will be painfully slow with ext3 on a 1TB or greater volume. XFS really is the only filesystem that was specificially designed for filesystems that big...and although others like JFS and ext3 will work I think that only XFS will provide adequate performance. Even though read-write performance isn't a factor here (as stated in the first post) I don't think the OP would want to spend minutes on end just waiting for a filesystem to mount or umount.

Though I've personally never used any volumes even close to that size, so I'm just having to base these performance estimates off of what I HAVE seen and what other people's benchmarks have shown...I could very well be wrong.

Edit: in addition to hokatichenci's post: Once again this goes to show that, at the end of the day, only XFS was designed to be used on volumes this large. No other file system can boast that.
 
Well, I have the time, I might just play around with each filesystem just for kicks (format using one, run tests, then reformat using the other and run more tests)... I looked them up on Wikipedia and it gave me some impressive information on both. I'm still leaning towards XFS, but I'm not going to rule out ext3 just yet.
 
You must log in or register to reply here.
Back
Top