UFS vs EXT3 vs EXT4 vs HAMMER vs NTFS vs zfs

alie

alie

hi guys,

which one is better ?
which one is faster and more stable ? is there any link that have comparisons for these file systems ?
 
It depends what the application is.

UFS has some features which were tacked on after the fact, like ACLs and journaling.

I'd bet that NTFS and ZFS are the best of your list, but I can't prove it.
 
It is very difficult to compare these side by side since they cannot all be used on one OS.
 
if i want use FreeBSD for Load balancing which one is good ?
if i want use FreeBSD for Desktop and do some programming which FS is good for this ?
 
If you want a desktop, you can safely go with any of these, UFS for example, no problems in 4 years of daily usage :)
 
i'm using zfs on fbsd-7.1-stable, great fs, you just need to talk to it, and is so simple, i love it..zfs is future, i'm so glad i can use it..thanks freebsd :))
 
but i heard there's UFS2, is it correct ? is zfs fully implemented for FreeBSD ?
 
I've been working for some time now with ZFS on OpenSolaris and I'm quite impressed by its performance and ease of use.
It's also excellent for making a cheap mirrored file server at home.
On the other hand for a daily user like me UFS performs fine.
 
You're right, but I think it ended without a clear answer. You know what, basically the question is so fundamental that it will never be outdated. Today, the discussion about ZFS and other filesystems is very heated; this is because of the new generation of SSDs.
 
You have to understand that there is no answer to this question. Is it better <yourdistro>Linux, FreeBSD, NetBSD, OpenBSD, OpenSolaris, OpenIndiana, Nexenta, Solaris, Darwin, .... ? It is the same question. There is no one-size-fits-all. At least this is my opinion.
Now, if you are using a pure FreeBSD environment, UFS2 or ZFS are the choice. The latter is really promising and is a good choice if you need it and have enough ram. The former is rock solid and combined with GEOM gives you a lot of power and control.
If you need to exchange data with other system, than a filesystem that is enough supported on both ends is the right choice (of course, if you cannot export a native filesystem with a networking one - CIFS or NFS mainly).
You have to consider what are your needs first, then what are your resources, and then you can choose the best filesystem.
I use UFS2 and ZFS day by day and I'm happy with both. I have also machines with ext3/ext4 and some USB media with ext2 and I'm happy also with those. My favourite remains ZFS, and I use whenever it is possible. NTFS is a forced choice for those media that are mounted also from Windows machine, but I don't consider it for anything else due to the license.
By the way, someone with much more experience than me can give you better hints.
 
Hi Fluca1978
You're relatively right but we are comparing file system with file system.
Of course we can.For instance Hammer is DragonflyBSD native file system.Hammer is very powerfull at infinite NFS-exportable snapshots, master-multislave operation, configurable history retention, fsckless-mount, and checksums to deal with data corruption.Also ZFS has same mechanism for data corruption that can be compaired and explaining all of it is beyond of this forum scope.
Every file system's have limitation's that must be consider.For example IOPS performance of a ZFS storage pool can suffer if the ZFS raid is not appropriately configured and this is not good for SSD fan's.A non-geek user doesn't care this feature's but everybody care about read/write speed and other visible factor's.
There is a very solid benchmark that may be useful:
http://www.phoronix.com/scan.php?page=article&item=dragonfly_hammer&num=2
 
RafaelAdorman said:
Hi Fluca1978
You're relatively right but we are comparing file system with file system.
Of course we can.
Click to expand...

I'm not saying that it is not correct to compare file systems. But to my extent it is not so useful, since we are talking about a comparisong between ntfs and ufs, ext3 and hammer. Now, as an example, ext3 is not native on freebsd, so it does not mean to me if it has a better design than ufs, since it will perform much more bad than the latter. The same is true for other comparisons.
Moreover, please consider that almost every filesystem of this thread subject has been deployed in server environment. Despite the final result, this means at least that the filesystem is performing not so badly, and therefore the right choice should depend on the environment itself. What are your purposes? What is your license? How are you going to exchange data? Which hardware are you using?
Is it very difficult to choose the right file system without having an application scenario. That is my idea.
Besides this, if you are running FreeBSD, go for both ZFS and UFS.
 
I think Hammer is a very interesting file system. It's a pity it doesn't have a volume manager.
The great advantage with Hammer is that it can rewrite data blocks. And that should in theory give a massive speedup using deduplication.
 
RafaelAdorman said:
Hi Fluca1978
You're relatively right but we are comparing file system with file system.
Of course we can.For instance Hammer is DragonflyBSD native file system.Hammer is very powerfull at infinite NFS-exportable snapshots, master-multislave operation, configurable history retention, fsckless-mount, and checksums to deal with data corruption.Also ZFS has same mechanism for data corruption that can be compaired and explaining all of it is beyond of this forum scope.
Every file system's have limitation's that must be consider.For example IOPS performance of a ZFS storage pool can suffer if the ZFS raid is not appropriately configured and this is not good for SSD fan's.A non-geek user doesn't care this feature's but everybody care about read/write speed and other visible factor's.
There is a very solid benchmark that may be useful:
http://www.phoronix.com/scan.php?page=article&item=dragonfly_hammer&num=2
Click to expand...

A RAID10 is always going to be best for IOPS regardless of what type of RAID you deploy. Has very little to do with the filesystem in question.

/Sebulon
 
AndyUKG said:
Sorry I don't understand your question/statement. The wiki supports what Sebulon said, RAID10 (or 1+0 however you want to write it) is best for IO for ZFS and as a general rule for any RAID system.

thanks Andy.
Click to expand...
Recently (post #15) i wrote something that was completely based on wiki.
Sebulon wrote something strange and based on something else(i think).
So i sent wiki's link as a proof.It's no brainer.
Beside if wiki support's what sebulon said then support's what i said.
So the question is: Which statement of wiki's link can prove your claim ?
 
RafaelAdorman said:
So the question is: Which statement of wiki's link can prove your claim ?
Click to expand...

Well I'm still not sure I understand you, are you referring to:

ZFS storage pool can suffer if the ZFS raid is not appropriately configured
Click to expand...

vs Sebulon and my assertion that on any RAID, RAID10 will always be faster?

The point is ZFS isn't just a file system, its a RAID system, volume manager and file system rolled into one. A file system that doesn't provide RAID has the same issues (performance depends on RAID configuration) when installed on top of a hardware or software RAID.
So, you can't put as a negative of Hammer or UFS, that the RAID has to be correctly configured, because Hammer and UFS don't provide RAID.

In reality if you want to compare like with like then you must compare ZFS vs Hammer on top of some RAID system.

thanks Andy.
 
No you can't compare ZFS vs Hammer on top of some raid system but you can compare same mechanism in both.Because ZFS has some capabilities that doesn't exist in Hammer and vice versa.
Technically;same category but not same philosophy.
For example you can compare "Backups and history" mechanism on both.
 
Well yes, yes you can compare them.

Anyway, I was trying to respond to your question, but as I suspected I don't seem to have understood it. If you can put it more clearly maybe I can answer it, but you are jumping around alot and not being specific in what you write in your posts...

My main point was that the wiki and Sebulon are in agreement, and I see no contradictory info in the wiki article...

thanks Andy.
 
AndyUKG said:
Well yes, yes you can compare them.

Anyway, I was trying to respond to your question, but as I suspected I don't seem to have understood it. If you can put it more clearly maybe I can answer it, but you are jumping around alot and not being specific in what you write in your posts...

My main point was that the wiki and Sebulon are in agreement, and I see no contradictory info in the wiki article...

thanks Andy.
Click to expand...
I think you and sebulon are in agreement(or you are same) my friend.
My post's are same but from different perspectives.
I know this is hard to accept but remember this is not about me and you.this is a concept and if you think you're true go ahead and modify wiki.
12 thanks in less than 11 minutes(every minute one hit) !!!
As i said billion times:"when you want to compare filesystem's you have to compare same mechanism's and same solutions and implementations for one problem"
 
You must log in or register to reply here.
Back
Top