I'm not totally sure about this, but I think Solaris 10 is not GPL. I think it's Sun's own open liscence.
Quote:
Originally Posted by blake.irvin
If they want deep enterprise penetration (which they haven't said yet), it would make some sense to at least consider putting a beefier server OS under OS X. Maybe I'm just engaging in some wishful thinking
Why would they bother porting ZFS to OS X if they were going to use Open Solaris? I've heard a lot of people mention a kernel switch but would the system even be binary compatible? If not, that idea is right out because we're not going through the whole update process again.
If the performance of 10.5 has improved to the extent we need then we don't need a new kernel as that is the only flaw in the current system. If ZFS offers IO improvements then this helps relieve a bottleneck too.
I agree with what was said about snapshots and Time Machine. It sounds to me like Time Machine was built purposely to take advantage of ZFS. This is the peak of OS X's development and what better way to go about it than to use a future-proof filesystem.
It also means that cross-platform issues are gone. If Solaris uses ZFS and Windows machines have to connect to ZFS-based servers then they'll have to be compatible, which makes Windows machines compatible with Macs and someone said Leopard will have NTFS read/write support so we're set.
I think the main concerns were that the filesystem is not optimized for a desktop OS but rather a server system but Apple can probably modify it enough to work better on a desktop. ZFS is sure starting to look like that secret feature (if there is only one).
As for CPU being eaten, they said that it uses minimal CPU resources and we have two CPUs anyway so we likely won't notice it happening.
About suboptimal block sizes, that other guy said it uses automatic block sizes, which sounds the most optimal in any situation. It allows you to specify your own block sizes if you want.
Backwards compatibility for ZFS is probably the only thing that concerns me but I'm sure there will be ways round it.
Just like the Intel switch, this may be painful at first but it should be for the better.
Why would they bother porting ZFS to OS X if they were going to use Open Solaris? I've heard a lot of people mention a kernel switch but would the system even be binary compatible? If not, that idea is right out because we're not going through the whole update process again.
If the performance of 10.5 has improved to the extent we need then we don't need a new kernel as that is the only flaw in the current system. If ZFS offers IO improvements then this helps relieve a bottleneck too.
I agree with what was said about snapshots and Time Machine. It sounds to me like Time Machine was built purposely to take advantage of ZFS. This is the peak of OS X's development and what better way to go about it than to use a future-proof filesystem.
It also means that cross-platform issues are gone. If Solaris uses ZFS and Windows machines have to connect to ZFS-based servers then they'll have to be compatible, which makes Windows machines compatible with Macs and someone said Leopard will have NTFS read/write support so we're set.
I think the main concerns were that the filesystem is not optimized for a desktop OS but rather a server system but Apple can probably modify it enough to work better on a desktop. ZFS is sure starting to look like that secret feature (if there is only one).
As for CPU being eaten, they said that it uses minimal CPU resources and we have two CPUs anyway so we likely won't notice it happening.
About suboptimal block sizes, that other guy said it uses automatic block sizes, which sounds the most optimal in any situation. It allows you to specify your own block sizes if you want.
Backwards compatibility for ZFS is probably the only thing that concerns me but I'm sure there will be ways round it.
Just like the Intel switch, this may be painful at first but it should be for the better.
While I'm only 50-50 about whether it will be THE new file format, rather than being an option as is the UNIX file format, I don't agree with that article either on several points.
The talk about the Solaris core is because the MACH core is not popular here, or on other sites. People are always looking for ways Apple would get rid of it.
All MACs now have at least two cores. Most tasks don't require two cores. Even when they do, they rarely use most of the power of the cores . That's rare.
Over time, most machines will have four cores. The loss of part of the performance of one will be missed even less. So I doubt that performance will be hit most of the time by this.
As far as backwards compatibility goes, he pointed out that this happened with System 8. I remember it well.
It doesn't support his argument. Rather, it supports the contrary one. That Apple has done that in the past, and could do it again, if they feel the advantages are worth it. It would also help to convince people to upgrade sooner.
I don't buy the block size argument either. Auto block size allocation would seem to eliminate that problem.
And as has been pointed out, Apple will no doubt fine tune the system for their needs. That's no doubt what at least some of those Apple engineers are doing.
Calm down, everybody. There won't be any ZFS by default in Leopard.
It is not trivial to implement! None of the builds of Leopard so far had ZFS as default - think about it. It's the exact same argument I used to predict that Leopard would be late - lack of "secret features" in the builds meant that Leopard would be late - since they *must* test them widely with developers before releasing. Exactly the same situation with ZFS - it ain't in the builds as the default file system, and you can't implement and test such a non-trivial feature so close to the release date (October). Ergo, no ZFS in Leopard by default. Period.
How big do filesystems need to be? In a world where 640KB is certainly not enough for computer memory, current filesystems have reached or are reaching the end of their usefulness. A 64-bit filesystem would meet today's need, but estimate of the lifetime of a 64-bit filesystem is about 10 years. Extending to 128-bits gives ZFS an expected lifetime of 30 years (UFS, for comparison, is about 20 years old). So how much data can you squeeze into a 128-bit filesystem? 16 exabytes or 18 million terabytes. How many files can you cram into a ZFS filesystem? 200 million million.
Could anyone use a fileystem that large? No, not really. The topic has roused discussions about boiling the oceans if a real life storage unit that size was powered on. It may not be necessary to have 128 bits, but it doesn't hurt and we won't have to worry about running out of addressable space.
Your math is wrong. 2^128 is 16EB^16EB, yes that's an exponent. 64bits only give you 16EB singularly.
How big is that? if you bought 16EB of RAM today it would approximately cost the GNP of the USA last year (~$2.6 Trillion if you want a ballpark figure). Now keep buying that every second for the next 16 BILLION YEARS. Yeah, we would run out of time with this sun, it would explode, a new sun would be born and we wouldn't even be halfway through with buying the RAM.
I highly doubt the human race can create that many files collectively in the next million years, let alone thhe next 30!
Sun has plans to license it under the GPLv3... or was that OpenSolaris? Either way pulling an entire kernel swap would be damn stupid right now, and as I said, Apple is going to do what works for them. What works for them right now is maintaining complete control of their own kernel, XNU, which is not a "server OS" and not even an OS itself, but a Kernel.
Calm down, everybody. There won't be any ZFS by default in Leopard.
It is not trivial to implement! None of the builds of Leopard so far had ZFS as default - think about it. It's the exact same argument I used to predict that Leopard would be late - lack of "secret features" in the builds meant that Leopard would be late - since they *must* test them widely with developers before releasing. Exactly the same situation with ZFS - it ain't in the builds as the default file system, and you can't implement and test such a non-trivial feature so close to the release date (October). Ergo, no ZFS in Leopard by default. Period.
I agree. There is no evidence that Apple is making ZFS the default filesystem. Except for blake.irvin's post (on page 2 of this thread) I've read absolutely nothing about successfully booting from ZFS.
I do see a lot of potential with ZFS but there is still much to do as far as I can tell. Plus, I expect Apple to implement ZFS in OS X Server first. After all, that is where this change would be the most beneficial.
The earliest we will see this as a default file system will be 10.6.
I agree. There is no evidence that Apple is making ZFS the default filesystem. Except for blake.irvin's post (on page 2 of this thread) I've read absolutely nothing about successfully booting from ZFS.
I do see a lot of potential with ZFS but there is still much to do as far as I can tell. Plus, I expect Apple to implement ZFS in OS X Server first. After all, that is where this change would be the most beneficial.
The earliest we will see this as a default file system will be 10.6.
I've been wondering if ZFS, and in specific its copy-on-write capabilities, aren't the foundation of Time Machine. However, perhaps for this to happen, ZFS wouldn't necessarily need to be the default FS of the boot partition, but rather the required FS of the external drive, share point, etc.
I've been wondering if ZFS, and in specific its copy-on-write capabilities, aren't the foundation of Time Machine.
What we do know...
Current Leopard builds can't enable ZFS.
Time Machine works on Leopard.
So we can deduce that Time Machine doesn't require ZFS for current Leopard builds. Now, I've read many reports that it's quite buggy. Perhaps the solution was to use ZFS instead of HFS+, but I doubt as it seems ZFS has enough of it's own problems at this time.
I really think ZFS is a long term goal for Apple, and not intended for Leopard.
but I doubt as it seems ZFS has enough of it's own problems at this time.
I really think ZFS is a long term goal for Apple, and not intended for Leopard.
Yeah it could easily go one way or the other but it's already in Solaris and I doubt Apple will be releasing 10.6 for a while and a filesystem change is even less likely to happen in a point release. So we'd be talking about 2 years maybe before we'd get to see the benefits. It definitely seems like ZFS is a superior file system to HFS+ and I don't think Apple ever like to be the ones playing catch up. The move would likely be more important for their Xserve business than their desktops but both would see benefits.
So we can deduce that Time Machine doesn't require ZFS for current Leopard builds. Now, I've read many reports that it's quite buggy. Perhaps the solution was to use ZFS instead of HFS+, but I doubt as it seems ZFS has enough of it's own problems at this time.
I really think ZFS is a long term goal for Apple, and not intended for Leopard.
(speculation on my part)
So are you saying Sun's CEO was lying the other day?
We're not going to have to wait for ZFS folks. Remember the WWDC beta is going to be about a full month later than the last Beta in early March. ZFS was workable two betas ago.
Sun is booting OpenSolaris on ZFS and Thumper uses it already production hardware.
So are you saying Sun's CEO was lying the other day?
I don't think he was lying, I just think that too mch is being read into the word "default." Don't get me wrong, ti would be great if Apple and Sun have worked out these ZFS issues but it seems like it's time prohibitive. I think the most likely answer is that he spoke incorrectly. He should have used option instead.
According to InfoWorld...
"Only hours after his initial posting, Sun's Hamilton revised his blog so that the ZFS line read: "Jonathan noted that Apple is planning to use the ZFS file system from OpenSolaris in future versions of their OS." In the blog comments, he also added: "I hope this clears up some of the confusion and concern I may have caused."
Though nice, it's just not the simplest answer. I gotta go with Hakim on this one, Baby!
Comments
If they want deep enterprise penetration (which they haven't said yet), it would make some sense to at least consider putting a beefier server OS under OS X. Maybe I'm just engaging in some wishful thinking
If the performance of 10.5 has improved to the extent we need then we don't need a new kernel as that is the only flaw in the current system. If ZFS offers IO improvements then this helps relieve a bottleneck too.
I agree with what was said about snapshots and Time Machine. It sounds to me like Time Machine was built purposely to take advantage of ZFS. This is the peak of OS X's development and what better way to go about it than to use a future-proof filesystem.
It also means that cross-platform issues are gone. If Solaris uses ZFS and Windows machines have to connect to ZFS-based servers then they'll have to be compatible, which makes Windows machines compatible with Macs and someone said Leopard will have NTFS read/write support so we're set.
Interesting article (not written by me)...
http://www.bynkii.com/archives/2007/..._file_sys.html
I think the main concerns were that the filesystem is not optimized for a desktop OS but rather a server system but Apple can probably modify it enough to work better on a desktop. ZFS is sure starting to look like that secret feature (if there is only one).
As for CPU being eaten, they said that it uses minimal CPU resources and we have two CPUs anyway so we likely won't notice it happening.
About suboptimal block sizes, that other guy said it uses automatic block sizes, which sounds the most optimal in any situation. It allows you to specify your own block sizes if you want.
Backwards compatibility for ZFS is probably the only thing that concerns me but I'm sure there will be ways round it.
Just like the Intel switch, this may be painful at first but it should be for the better.
Why would they bother porting ZFS to OS X if they were going to use Open Solaris? I've heard a lot of people mention a kernel switch but would the system even be binary compatible? If not, that idea is right out because we're not going through the whole update process again.
If the performance of 10.5 has improved to the extent we need then we don't need a new kernel as that is the only flaw in the current system. If ZFS offers IO improvements then this helps relieve a bottleneck too.
I agree with what was said about snapshots and Time Machine. It sounds to me like Time Machine was built purposely to take advantage of ZFS. This is the peak of OS X's development and what better way to go about it than to use a future-proof filesystem.
It also means that cross-platform issues are gone. If Solaris uses ZFS and Windows machines have to connect to ZFS-based servers then they'll have to be compatible, which makes Windows machines compatible with Macs and someone said Leopard will have NTFS read/write support so we're set.
I think the main concerns were that the filesystem is not optimized for a desktop OS but rather a server system but Apple can probably modify it enough to work better on a desktop. ZFS is sure starting to look like that secret feature (if there is only one).
As for CPU being eaten, they said that it uses minimal CPU resources and we have two CPUs anyway so we likely won't notice it happening.
About suboptimal block sizes, that other guy said it uses automatic block sizes, which sounds the most optimal in any situation. It allows you to specify your own block sizes if you want.
Backwards compatibility for ZFS is probably the only thing that concerns me but I'm sure there will be ways round it.
Just like the Intel switch, this may be painful at first but it should be for the better.
While I'm only 50-50 about whether it will be THE new file format, rather than being an option as is the UNIX file format, I don't agree with that article either on several points.
The talk about the Solaris core is because the MACH core is not popular here, or on other sites. People are always looking for ways Apple would get rid of it.
All MACs now have at least two cores. Most tasks don't require two cores. Even when they do, they rarely use most of the power of the cores . That's rare.
Over time, most machines will have four cores. The loss of part of the performance of one will be missed even less. So I doubt that performance will be hit most of the time by this.
As far as backwards compatibility goes, he pointed out that this happened with System 8. I remember it well.
It doesn't support his argument. Rather, it supports the contrary one. That Apple has done that in the past, and could do it again, if they feel the advantages are worth it. It would also help to convince people to upgrade sooner.
I don't buy the block size argument either. Auto block size allocation would seem to eliminate that problem.
And as has been pointed out, Apple will no doubt fine tune the system for their needs. That's no doubt what at least some of those Apple engineers are doing.
It is not trivial to implement! None of the builds of Leopard so far had ZFS as default - think about it. It's the exact same argument I used to predict that Leopard would be late - lack of "secret features" in the builds meant that Leopard would be late - since they *must* test them widely with developers before releasing. Exactly the same situation with ZFS - it ain't in the builds as the default file system, and you can't implement and test such a non-trivial feature so close to the release date (October). Ergo, no ZFS in Leopard by default. Period.
2. Honkin' big filesystems
How big do filesystems need to be? In a world where 640KB is certainly not enough for computer memory, current filesystems have reached or are reaching the end of their usefulness. A 64-bit filesystem would meet today's need, but estimate of the lifetime of a 64-bit filesystem is about 10 years. Extending to 128-bits gives ZFS an expected lifetime of 30 years (UFS, for comparison, is about 20 years old). So how much data can you squeeze into a 128-bit filesystem? 16 exabytes or 18 million terabytes. How many files can you cram into a ZFS filesystem? 200 million million.
Could anyone use a fileystem that large? No, not really. The topic has roused discussions about boiling the oceans if a real life storage unit that size was powered on. It may not be necessary to have 128 bits, but it doesn't hurt and we won't have to worry about running out of addressable space.
Your math is wrong. 2^128 is 16EB^16EB, yes that's an exponent. 64bits only give you 16EB singularly.
How big is that? if you bought 16EB of RAM today it would approximately cost the GNP of the USA last year (~$2.6 Trillion if you want a ballpark figure). Now keep buying that every second for the next 16 BILLION YEARS. Yeah, we would run out of time with this sun, it would explode, a new sun would be born and we wouldn't even be halfway through with buying the RAM.
I highly doubt the human race can create that many files collectively in the next million years, let alone thhe next 30!
Sun has plans to license it under the GPLv3... or was that OpenSolaris? Either way pulling an entire kernel swap would be damn stupid right now, and as I said, Apple is going to do what works for them. What works for them right now is maintaining complete control of their own kernel, XNU, which is not a "server OS" and not even an OS itself, but a Kernel.
Sebastian
http://www.opensolaris.org/jive/thre...23699&tstart=0
o GPL* licensing OpenSolaris would be yielding to a small vocal
minority of FOSS developers who use the lack of GPL licensing, purely
as a means of fostering FUD towards OpenSolaris and who will, in all
likelyhood, find some other workable mechanism to continue to foster
FUD towards the project.
Kudos to the OGB. Dang tired of the FSF and GPL v3...
Vinea
Will this new file system require a new harddrive or new hardware, or would it work with my current (June 2006) MacBook?
Nope, not at all.
Nope, not at all.
ummm.... That was not a yes or no question...
Calm down, everybody. There won't be any ZFS by default in Leopard.
It is not trivial to implement! None of the builds of Leopard so far had ZFS as default - think about it. It's the exact same argument I used to predict that Leopard would be late - lack of "secret features" in the builds meant that Leopard would be late - since they *must* test them widely with developers before releasing. Exactly the same situation with ZFS - it ain't in the builds as the default file system, and you can't implement and test such a non-trivial feature so close to the release date (October). Ergo, no ZFS in Leopard by default. Period.
I agree. There is no evidence that Apple is making ZFS the default filesystem. Except for blake.irvin's post (on page 2 of this thread) I've read absolutely nothing about successfully booting from ZFS.
I do see a lot of potential with ZFS but there is still much to do as far as I can tell. Plus, I expect Apple to implement ZFS in OS X Server first. After all, that is where this change would be the most beneficial.
The earliest we will see this as a default file system will be 10.6.
ummm.... That was not a yes or no question...
It was a poorly structured, dichotomous question. I responded to the relevant first part as the latter part was redundant.
"Will this new file system require a new harddrive or new hardware?"
"Nope, not at all."
I agree. There is no evidence that Apple is making ZFS the default filesystem. Except for blake.irvin's post (on page 2 of this thread) I've read absolutely nothing about successfully booting from ZFS.
I do see a lot of potential with ZFS but there is still much to do as far as I can tell. Plus, I expect Apple to implement ZFS in OS X Server first. After all, that is where this change would be the most beneficial.
The earliest we will see this as a default file system will be 10.6.
I've been wondering if ZFS, and in specific its copy-on-write capabilities, aren't the foundation of Time Machine. However, perhaps for this to happen, ZFS wouldn't necessarily need to be the default FS of the boot partition, but rather the required FS of the external drive, share point, etc.
Just a thought
I've been wondering if ZFS, and in specific its copy-on-write capabilities, aren't the foundation of Time Machine.
What we do know...
- Current Leopard builds can't enable ZFS.
- Time Machine works on Leopard.
So we can deduce that Time Machine doesn't require ZFS for current Leopard builds. Now, I've read many reports that it's quite buggy. Perhaps the solution was to use ZFS instead of HFS+, but I doubt as it seems ZFS has enough of it's own problems at this time.I really think ZFS is a long term goal for Apple, and not intended for Leopard.
(speculation on my part)
but I doubt as it seems ZFS has enough of it's own problems at this time.
I really think ZFS is a long term goal for Apple, and not intended for Leopard.
Yeah it could easily go one way or the other but it's already in Solaris and I doubt Apple will be releasing 10.6 for a while and a filesystem change is even less likely to happen in a point release. So we'd be talking about 2 years maybe before we'd get to see the benefits. It definitely seems like ZFS is a superior file system to HFS+ and I don't think Apple ever like to be the ones playing catch up. The move would likely be more important for their Xserve business than their desktops but both would see benefits.
What we do know...
- Current Leopard builds can't enable ZFS.
- Time Machine works on Leopard.
So we can deduce that Time Machine doesn't require ZFS for current Leopard builds. Now, I've read many reports that it's quite buggy. Perhaps the solution was to use ZFS instead of HFS+, but I doubt as it seems ZFS has enough of it's own problems at this time.I really think ZFS is a long term goal for Apple, and not intended for Leopard.
(speculation on my part)
So are you saying Sun's CEO was lying the other day?
Sun is booting OpenSolaris on ZFS and Thumper uses it already production hardware.
case insensitivity has been "fast tracked"
http://www.opensolaris.org/os/commun...C4644C98986C1F
With 4 months to sweeten the Leopard Beta before October I have no doubts that ZFS will be ready in October. Simply give users the choice.
1. Install HFS+ (Recommended)
2. Install ZFS (read help file about concerns)
I'd rather run Time Machine on ZFS myself.
So are you saying Sun's CEO was lying the other day?
I don't think he was lying, I just think that too mch is being read into the word "default." Don't get me wrong, ti would be great if Apple and Sun have worked out these ZFS issues but it seems like it's time prohibitive. I think the most likely answer is that he spoke incorrectly. He should have used option instead.
According to InfoWorld...
"Only hours after his initial posting, Sun's Hamilton revised his blog so that the ZFS line read: "Jonathan noted that Apple is planning to use the ZFS file system from OpenSolaris in future versions of their OS." In the blog comments, he also added: "I hope this clears up some of the confusion and concern I may have caused."
Though nice, it's just not the simplest answer. I gotta go with Hakim on this one, Baby!
ummm.... That was not a yes or no question...
It wasn't a yes or no answer either. Was he answering yes, or no?
So are you saying Sun's CEO was lying the other day?
I don't like these kinds of responses.
If you mean; was he mistaken?, that would be more correct.