Linked by Thom Holwerda on Wed 4th Nov 2009 16:48 UTC
Linux Linux continues to innovate in the area of file systems. It supports the largest variety of file systems of any operating system. It also provides cutting-edge file system technology. Two new file systems that are making their way into Linux include the NiLFS(2) log-structured file system and the exofs object-based storage system. Discover the purpose behind these two new file systems and the advantages that they bring.
Order by: Score:
Why all the filesystems.
by theTSF on Wed 4th Nov 2009 17:11 UTC
theTSF
Member since:
2005-09-27

This is something I never really get about Linux... Its fascination of Filesystems.

I can see a new modern file system every 10 years or so. However most other OS's put there efforts elsewhere. I think Linux should too. I think it is a case that File Systems are fun to design and program. However there are areas that are much more annoying to design and code that needs to be addressed.

Reply Score: 1

RE: Why all the filesystems.
by TechGeek on Wed 4th Nov 2009 17:19 UTC in reply to "Why all the filesystems."
TechGeek Member since:
2006-01-14

You miss the point that filesystems are actually an important part of any OS. Different filesystems do different things well. And really, most of the filesystems in Linux are there for compatibility with other systems. Most distros use the ext series. Besides btrfs, what other filesystems were created explicitly for Linux?

Reply Score: 2

RE[2]: Why all the filesystems.
by Tsukasa on Wed 4th Nov 2009 18:05 UTC in reply to "RE: Why all the filesystems."
Tsukasa Member since:
2006-05-15

Besides btrfs, what other filesystems were created explicitly for Linux?

reiserfs comes to my mind.

I do agree that having a plethora of file-systems for compatibility is great, even when they only allow read access.

Reply Score: 1

RE: Why all the filesystems.
by Praxis on Wed 4th Nov 2009 17:27 UTC in reply to "Why all the filesystems."
Praxis Member since:
2009-09-17

well behind most of these file systems are some pretty major companies, I think its safe to say that Oracle and IBM aren't just developing these things for the heck of it. One thing to keep in mind is that it takes a long time to properly develop a file system that people will trust with their data, so you need to start creating these things about 5-7 years before you think you will need it.

Reply Score: 2

RE: Why all the filesystems.
by Delgarde on Wed 4th Nov 2009 20:44 UTC in reply to "Why all the filesystems."
Delgarde Member since:
2008-08-19

I can see a new modern file system every 10 years or so. However most other OS's put there efforts elsewhere. I think Linux should too. I think it is a case that File Systems are fun to design and program. However there are areas that are much more annoying to design and code that needs to be addressed.


Right, but what do you mean, "Linux should put it's efforts elsewhere"? Linux isn't a person or organisation - it's a piece of software developed by a bunch of individuals all working for their own (or their employers) different purposes.

Reading the article, these two filesystems appear to have been contributed by companies who presumably needed new filesystems for whatever reason. So it's not exactly productive to tell those people they should have been working on something else instead.

For better or worse, that's the open-source way - people work on whatever they want to work on. It's what you get, when all of your contributors are volunteers.

Reply Score: 5

v RE[2]: Why all the filesystems.
by strcpy on Wed 4th Nov 2009 20:50 UTC in reply to "RE: Why all the filesystems."
RE[3]: Why all the filesystems.
by segedunum on Wed 4th Nov 2009 21:13 UTC in reply to "RE[2]: Why all the filesystems."
segedunum Member since:
2005-07-06

I'm afraid there is no disconnect between the two statements in the way that you want.

A lot of different companies contribute developers and resources to the Linux kernel and they each work on things that make sense to them, so yes, they do volunteer for projects. They cannot be forced to work on anything by any central authority.

You've tried to argue this line before, and it usually boils down to you trying to insinuate that open source development is turning to some central proprietary development model simply because some people happen to be getting paid to write open source code within the most important projects. Not only is it wrong, it's also an argument that goes absolutely nowhere.

Reply Score: 4

RE[4]: Why all the filesystems.
by strcpy on Thu 5th Nov 2009 05:04 UTC in reply to "RE[3]: Why all the filesystems."
strcpy Member since:
2009-05-20


You've tried to argue this line before, and it usually boils down to you trying to insinuate that open source development is turning to some central proprietary development model simply because some people happen to be getting paid to write open source code within the most important projects. Not only is it wrong, it's also an argument that goes absolutely nowhere.


Good that you noticed the overall point. I still try to argue on those lines, albeit my argument may not be that coherent.

The central point is that open source development in resembling more and more closed development, not because people get paid to write open source code, but because the agenda and weight that the big parties carry when decisions are made. Sad bad quite true.

Why is it wrong, if I may ask?

EDIT: typos.

Edited 2009-11-05 05:11 UTC

Reply Score: 2

RE[5]: Why all the filesystems.
by gustl on Thu 5th Nov 2009 17:25 UTC in reply to "RE[4]: Why all the filesystems."
gustl Member since:
2006-01-19

Especially where Linux (the kernel) is concerned, I cannot see a single company being able to buy any of Linus' decisions.

Not because Linus is beyond accepting money for a favour, but because all companies know that forks would spring up like mushrooms the day it happened.

None of the companies wants to further develop Linux on their own. Even Google rebases their internal Linux fork onto a mainline kernel from time to time.

Reply Score: 3

RE: Why all the filesystems.
by glarepate on Fri 6th Nov 2009 04:54 UTC in reply to "Why all the filesystems."
glarepate Member since:
2006-01-04

However there are areas that are much more annoying to design and code that needs to be addressed.

And who should be assigned to "fix" those areas?

I'm presuming that you have other things to do and see your contribution as pointing out the need for work on the [unspecified] design and code areas.

How will you ensure that those important areas are addressed, much less addressed effectively and sustainably? Do you have a suggestion on how someone else might accomplish or even supervise those assigned to the needed work? [In my own mind] I'm not complaining about you complaining, just wondering what you see as solutions to the problem of people working on what they like rather than what may be essential to the over all good of the OS.

Reply Score: 2

Comment by sysconfig
by sysconfig on Wed 4th Nov 2009 17:24 UTC
sysconfig
Member since:
2009-07-21

The only filesystem, which I would consider truly innovative and useful is ZFS. And that has not even made it to Linux yet (various reasons).

Filesystems nowadays need the ability to grow or shrink on the fly, hot snapshots, building/moving/expanding or even dissolving RAIDs etc.
And that has to happen while still being online and mounted.

Reply Score: 6

RE: Comment by sysconfig
by WereCatf on Wed 4th Nov 2009 17:47 UTC in reply to "Comment by sysconfig"
WereCatf Member since:
2006-02-15

The only filesystem, which I would consider truly innovative and useful is ZFS. And that has not even made it to Linux yet (various reasons).

I agree that ZFS is both innovative and useful and it would have been great if it were possible to include it in Linux, but as that is not possible there's coming up another fs to use: Btrfs. It seems it has more-or-less all the capabilities of ZFS and a few of its own new ones.

No, it's not the same thing, but it's close ;) Unfortunately it's not yet ready for consumption, but hopefully soon.

For those who are interested: http://btrfs.wiki.kernel.org/index.php/Main_Page

Reply Score: 4

RE: Comment by sysconfig
by akro on Wed 4th Nov 2009 17:48 UTC in reply to "Comment by sysconfig"
akro Member since:
2005-07-06

You do realize ADV-FS did most of those things over 10 years ago...

http://en.wikipedia.org/wiki/AdvFS

Reply Score: 2

v Comment by Oliver
by Oliver on Wed 4th Nov 2009 17:37 UTC
RE: Comment by Oliver
by merkoth on Wed 4th Nov 2009 17:51 UTC in reply to "Comment by Oliver"
merkoth Member since:
2006-09-22

May I ask why is so? Or should I refrain from feeding the troll?

Reply Score: 2

RE[2]: Comment by Oliver
by sbenitezb on Wed 4th Nov 2009 19:25 UTC in reply to "RE: Comment by Oliver"
sbenitezb Member since:
2005-07-22

Mmm... but it's true. The only innovation was to come from reiser4 forever. Btrfs is just Oracle attempt to bring ZFS-like FS to Linux kernel. Ext4 is Ext3++. The rest are domain specific FS. Too little innovation. But then, innovation is just a buzzword.

Reply Score: 2

RE[3]: Comment by Oliver
by strcpy on Wed 4th Nov 2009 20:16 UTC in reply to "RE[2]: Comment by Oliver"
strcpy Member since:
2009-05-20

Mmm... but it's true. The only innovation was to come from reiser4 forever. Btrfs is just Oracle attempt to bring ZFS-like FS to Linux kernel. Ext4 is Ext3++. The rest are domain specific FS. Too little innovation. But then, innovation is just a buzzword.


And Ext4 is based on Ext3 is based on Ext2 is based on UFS.

Reply Score: 2

RE[3]: Comment by Oliver
by kragil on Wed 4th Nov 2009 21:41 UTC in reply to "RE[2]: Comment by Oliver"
kragil Member since:
2006-01-04

http://lwn.net/Articles/342892/

An informed outsider might consider BtrFS as just a ZFS clone, but if you look at the implementation it is a vastly different and more modern beast.

Reply Score: 2

RE[4]: Comment by Oliver
by strcpy on Thu 5th Nov 2009 05:09 UTC in reply to "RE[3]: Comment by Oliver"
strcpy Member since:
2009-05-20

Yeah. Its teh best. And using hyperlinks as an argument is the new way of pushing your argument when you are too tired to write anything.

Edited 2009-11-05 05:10 UTC

Reply Score: 0

RE: Comment by Oliver
by jgagnon on Wed 4th Nov 2009 19:02 UTC in reply to "Comment by Oliver"
jgagnon Member since:
2008-06-24

Perhaps you should pick up a dictionary and look up that word "innovation". You might be surprised at its meaning.

Reply Score: 1

Competition is good
by Kebabbert on Wed 4th Nov 2009 22:40 UTC
Kebabbert
Member since:
2007-07-27

Ive read the article and I like it. I only like the best and coolest technology, that is why I like SUN best. ZFS, DTrace, Zones, Niagara, etc.

But for these filesystems, I would love to see them in production. The only problem is that it takes at least 6-7 years to develop a filesystem. It is much more tricky than develop a kernel. If you mess up the kernel, you loose a couple of hours of work, maybe. If you mess up the filesystem, you can loose several years of work. Filesystems are the nervous system, the skeleton of OS. Therefore I doubt BTRFS will make it on time. It looks like a good ZFS copy, but it will take several years more for it to be used in production. Even ZFS has bugs, 7-8 years later. And everyone agrees that SUNs engineers are very good and innovative, they produce hot tech that everyone wants.

But, I would like to see these filesystems to happen. It would only make ZFS better, becuase competition is good for everyone. The products get better, the users will benefit. And if these filesystems turn out to be better than ZFS, I will switch immediately. Why, I know OpenSolaris, it is very similar to Linux. It costs me almost nothing to relearn. OTOH, To go from Win7 to OpenVMS would take much effort.

Reply Score: 1

RE: Competition is good
by kaiwai on Thu 5th Nov 2009 02:09 UTC in reply to "Competition is good"
kaiwai Member since:
2005-07-06

Ive read the article and I like it. I only like the best and coolest technology, that is why I like SUN best. ZFS, DTrace, Zones, Niagara, etc.

But for these filesystems, I would love to see them in production. The only problem is that it takes at least 6-7 years to develop a filesystem. It is much more tricky than develop a kernel. If you mess up the kernel, you loose a couple of hours of work, maybe. If you mess up the filesystem, you can loose several years of work. Filesystems are the nervous system, the skeleton of OS. Therefore I doubt BTRFS will make it on time. It looks like a good ZFS copy, but it will take several years more for it to be used in production. Even ZFS has bugs, 7-8 years later. And everyone agrees that SUNs engineers are very good and innovative, they produce hot tech that everyone wants.

But, I would like to see these filesystems to happen. It would only make ZFS better, becuase competition is good for everyone. The products get better, the users will benefit. And if these filesystems turn out to be better than ZFS, I will switch immediately. Why, I know OpenSolaris, it is very similar to Linux. It costs me almost nothing to relearn. OTOH, To go from Win7 to OpenVMS would take much effort.


The problem I have with OpenSolaris is the horrible hardware support, for example, Broadcom 43XX support (from Broadcom themselves) is out in the form of a operating system agnostic hybrid driver and yet has not be ported to OpenSolaris by Sun programmers - one of the most common wireless chipsets out there along with the Intel wireless chipsets. Then there is the issue of HAL being replaced and Sun programmers being caught with their pants down - what are they putting up as a replacement? I see no open discussion about the replacement beyond a couple of people putting out the idea of using a combination of libdevinfo and libsysevent but nothing has developed beyond that. Then there is the issue of making the kernel ticklesss but next to nothing has been done of any sizeable degree to push down power usage and improve battery life on laptops.

If you're talking about OpenSolaris in the capacity of a server - no problems but out side of a large scale server OpenSolaris leaves a lot to be desired. What Linux has is the ability to scale from laptop to desktop to server - I simply don't see that happening in the OpenSolaris world; I see an operating system hobbling along struggling to remain relevant when all the competition has zoomed ahead in leaps and bounds.

Edited 2009-11-05 02:18 UTC

Reply Score: 3

RE[2]: Competition is good
by strcpy on Thu 5th Nov 2009 05:09 UTC in reply to "RE: Competition is good"
strcpy Member since:
2009-05-20


The problem I have with OpenSolaris is the horrible hardware support, for example, Broadcom 43XX support (from Broadcom themselves) is out in the form of a operating system agnostic hybrid driver and yet has not be ported to OpenSolaris by Sun programmers - one of the most common wireless chipsets out there along with the Intel wireless chipsets. Then there is the issue of HAL being replaced and Sun programmers being caught with their pants down - what are they putting up as a replacement?


This is getting a little tiresome. Linux has better hardware support, and everyone knows that. No doubt about it. But if you want to play that game, you can easily also turn it around and say that Windows is the best of them all because it clearly has the best hardware support for consumer PCs. (Drivers may not perhaps directly in kernel, but still every single gadget is supported in Windows.)


If you're talking about OpenSolaris in the capacity of a server - no problems but out side of a large scale server OpenSolaris leaves a lot to be desired. What Linux has is the ability to scale from laptop to desktop to server - I simply don't see that happening in the OpenSolaris world; I see an operating system hobbling along struggling to remain relevant when all the competition has zoomed ahead in leaps and bounds.



I'd imagine OpenSolaris is mostly used as a small or large server. This is what you also propose, then start to make claims about possible problems of scaling to a laptop? Just does not compute.

Reply Score: 1

RE[2]: Competition is good
by Kebabbert on Thu 5th Nov 2009 08:35 UTC in reply to "RE: Competition is good"
Kebabbert Member since:
2007-07-27

"The problem I have with OpenSolaris is the horrible hardware support, for example, Broadcom 43XX support (from Broadcom themselves) is out in the form of a operating system agnostic hybrid driver and yet has not be ported to OpenSolaris by Sun programmers - one of the most common wireless chipsets out there along with the Intel wireless chipsets. Then there is the issue of HAL being replaced and Sun programmers being caught with their pants down - what are they putting up as a replacement? I see no open discussion about the replacement beyond a couple of people putting out the idea of using a combination of libdevinfo and libsysevent but nothing has developed beyond that. Then there is the issue of making the kernel ticklesss but next to nothing has been done of any sizeable degree to push down power usage and improve battery life on laptops."

Yes, it is a problem that OpenSolaris has not as many drivers as Linux has. This makes the usability experience bad.

The other problems you mention, I dont see them as big problems. You mention a few problems, but there are other problems that other OSes has. Which you dont mention. For instance Linux broken sound API, lack stable ABI, etc. IMO, that is a bigger problem than those you mentioned.



"If you're talking about OpenSolaris in the capacity of a server - no problems but out side of a large scale server OpenSolaris leaves a lot to be desired. What Linux has is the ability to scale from laptop to desktop to server - I simply don't see that happening in the OpenSolaris world; I see an operating system hobbling along struggling to remain relevant when all the competition has zoomed ahead in leaps and bounds."

Yes, Linux has that ability to scale from laptop to server, but how good is Linux ability? As a server it sucks, compared to a real Unix or Mainframe. It is bloated and unstable. As a desktop it sucks compared to Windows or Mac OS X, in terms of usability. On the laptop, it might be good, I dont know. Jack of all trades, master of none.

OpenSolaris also scales from Asus EEE pc up to Big Iron. With the very same kernel. That is scalability. Linux has different kernels for different needs. You have to recompile it. Not scalable, but modifiable.

Regarding "the competition zoomed ahead in leaps and bounds" - I dont agree with you. Yes, the competition is ahead in terms of driver support. But if the nr of drivers is your criterion for an OS you want to use, good for you. Other people prefer stability and hot technology such as ZFS, DTrace - and they are easy to use too. There are numerous of more important things where OpenSolaris is ahead all other OSes. It is just a matter of OpenSolaris catching up on the user side. That is not that hard to do, compared to develop new hot killer tech. It is "just" porting/writing drivers, which any idiot can do. And then OpenSolaris continues to copy Ubuntu in terms of useability and GUI. Even now OpenSolaris is very similar to Ubuntu, almost a true copy. OpenSolaris benefits of all programs developed for Linux, just port them easily.

Which is more difficult? Develop hot new unique tech which no OS ever had, or copy other OSes useability and write drivers? I prefer that my OS just has to copy another OS gui a few years, than develop new revolutionizing tech. It is more difficult to copy new hot tech.

Reply Score: 3

SUN needs to be plagued
by eydaimon on Thu 5th Nov 2009 17:11 UTC
eydaimon
Member since:
2006-03-22

SUN needs to be plagued into opening ZFS licensing and make it a mainstream FS. Recent indication is that because of licensing issues, OS X did not get ZFS. That's a real shame.

Reply Score: 1

RE: SUN needs to be plagued
by Kebabbert on Fri 6th Nov 2009 20:33 UTC in reply to "SUN needs to be plagued"
Kebabbert Member since:
2007-07-27

The reason Mac OS X didnt get ZFS is that Apple wanted to have some special treatment from SUN. Special license, etc. SUN couldnt do that, on Apple's terms.

This is what Jeff Bonwick, the main ZFS architect, wrote on the OpenSolaris forum. Look for the thread... "Apple wont get ZFS" or something.

And I dont get it, if OpenSolaris is so bad and slow and inferior to Linux and all other OS, why do everyone want (and copy) ZFS and DTrace? Is OpenSolaris a piece of shit, or is it good tech? Can you guys make up your mind??

Reply Score: 2