What's next after HFS+?

[hobowankenobi]

Posting here cuz it is usually storage and server folks who know/care most:

Anybody following along remembers some years back when Apple was expected to move from HFS+ to ZFS. They even had a public beta at one point.

That project died...likely for both technical and licensing reasons.

Data sets continue to grow. Yes, Apple is no longer in the enterprise hardware biz, but it is not uncommon for "pro users" to have many TBs of data, and there is no end in sight with RAW photo and HD video space and performance needs.

So, the question is.....what is next, and how far can they stretch HFS+?

File Systems are way above my pay grade, but smarter folks are concerned about data integrity and loads of enterprise features....that HFS+ and other legacy file systems don't/can't implement.

ZFS seems to gaining traction in the storage world, if only very slowly, with some challenges that keep it from being anywhere close to easy-to-use in an Apple sort of way. Several projects have started and gone by the way side to have a Mac friendly option. OpenZFSonOSX looks hopeful, but it seems to be moving slowly...and it may die on the vine like previous attempts.

What really got me thinking about this though...having watched and waited patiently for so many years, was when I noticed that Synology appears to be rolling out Btrfs for their 6.0 OS beta:

What’s New in DSM 6.0 Beta 1

1. Btrfs File System
   • Set usage quota and user quota on shared folders.
   • Enable file compression for more efficient volume usage.
   • Perform defragmentation to increase system performance.
   • Data checksum is enabled by default and data scrubbing can be run to verify data integrity.
   • When file versioning is enabled in Cloud Station, it will not occupy additional storage, which would be the same amount of size as each shared folder.
   • Files will be backed up at the point in time when a backup task is started, even if the files are changed or deleted afterwards.
   • It is not supported to convert existing ext4 file system to Btrfs. You need to create a new one and move files on your own.

So....if a relatively small company like Synology can roll out a a modern file system like Btrfs, and have it bootable, customized, and running on very low power hardware resources (CPU/RAM), when will Apple deliver something similar?

 

[cruisin]

If you look at their Pro line (both trashcan and notebook and maybe the iPad) they do not support large amounts of internal storage, no where near the many TBs of data you refer to. They expect you to have an external storage array that plugs into your Mac or even is shared wirelessly. So if everyone will have massive external storage pretending to be a logical volume, why would the Mac need ZFS? I think they realized that the ZFS issues can be safely called somebody else's problem and left it at that.

Maybe if they kept the Mac Server and the corresponding Xserve raid they would actually try for ZFS, but they got rid of that product years ago.

 

[ChrisA]

Apple seems to be thinking that all data shall live in the cloud. That cloud might be a server in you office or it might be Apple's iCloud or whatever. With storage on the network and Mac internal storage limited to a couple TB at most or more likely less than 1TB a file system like ZFS is not needed

If you want ZFS then best way to get it is to get Solars. The second best way is to get BSD

I was disappointed when Oracle bought Sun. I wanted Apple to buy Sun and get into enterprises computing. maybe even keeping the Sun brand name for that.

ZFS is great if you have many, many TB of data on dozens of hard drives and a professional staff. I can't see the value of using ZFS for small systems.

 

[hobowankenobi]

Apple seems to be thinking that all data shall live in the cloud. That cloud might be a server in you office or it might be Apple's iCloud or whatever. With storage on the network and Mac internal storage limited to a couple TB at most or more likely less than 1TB a file system like ZFS is not needed

If you want ZFS then best way to get it is to get Solars. The second best way is to get BSD

I was disappointed when Oracle bought Sun. I wanted Apple to buy Sun and get into enterprises computing. maybe even keeping the Sun brand name for that.

ZFS is great if you have many, many TB of data on dozens of hard drives and a professional staff. I can't see the value of using ZFS for small systems.

I too hoped that Apple would buy Sun. Could have been a good fit.

I tend to agree that ZFS will never come to Apple. But that question stands....what will come after HFS+ and when?

 

[hobowankenobi]

If you look at their Pro line (both trashcan and notebook and maybe the iPad) they do not support large amounts of internal storage, no where near the many TBs of data you refer to. They expect you to have an external storage array that plugs into your Mac or even is shared wirelessly. So if everyone will have massive external storage pretending to be a logical volume, why would the Mac need ZFS? I think they realized that the ZFS issues can be safely called somebody else's problem and left it at that.

Maybe if they kept the Mac Server and the corresponding Xserve raid they would actually try for ZFS, but they got rid of that product years ago.

Agreed, Apple has no storage products, and none likely coming. But, as you say, Apple would point you at a Promise or G-Tech RAID box...that could be up to 64TB. At the rate of HD improvements and storage growth, we could be pushing 100TB HFS arrays in a couple years.

How would that be formatted and used by a pro user or small business owner? Right now....we know it is gonna be HFS via Disk Utility 99+% of the time. There is no other realistic option outside of the enterprise world.

Most experts agree that such large a storage array needs a better file system: more reliable, more efficient, snap shot-able, self healing, etc. Surely we can all agree that since end users can easily buy recommended hardware that is plug and play with such large amount of storage, the need for a consumer friendly modern file system is coming quick. I would suggest it is here now, and delivery is already late.

The question is not if, but when we will see it. The fact that ZFS got to open beta so many years ago shows there was interest, need, and assets allocated to file system improvements....well before we would dream of dozens TB of thunderbolt storage on users desk that we see today. The harder part to guess is will they try to stretch HFS again somehow as they have over the years, or replace it?

Replacing it is a big challenge, not for new machines and new users, but updates, upgrades, etc. Wiping a boot drive and formatting it is not something that they would want to suggest to the average user. But if it were a non-booting, storage only format, then there would be no need to "upgrade" any existing machine boot OSes. Heck, we have that now in DU, with several other format options, most of which average Mac users will never need. Adding a needed one can't be too revolutionary.....

OTOH, I was working IT on Macs when we had to switch from APM to GUID, and yes, we had to reformat about 75 machines to be able to do firmware updates, so this sort of disruptive change is not unheard of.

And let's not forget the doom and gloom predictions with the rumor of moving from PPC to X86 started appearing. Challenging, yes. Undoable...no. This, in comparison, is a much smaller challenge.

 

[Beachguy]

HFS++? HFS#? Objective-HFS?

Sorry... It's Friday and I couldn't resist being silly for a moment.

 

[2984839]

ZFS is outrageously complex. HAMMER2 offers similar features with much less code and would be a better choice to port to OS X, both for simplicity and licensing reasons.

 

[hobowankenobi]

Silly is good.

Yeah, agreed: likely not ZFS...from what I read.

Knowing Apple, and with the license issues they had with open source ZFS, I expect proprietary. I just hope we see something soon....

 

[Pakaku]

I don't think Apple ever intends to take mass local-storage seriously. You can see it in their iPhones and iPads because, price aside, they can get away with tiny SSDs. You can also infer it from the platter drives in their iMacs. You can also infer it from how they've cut out so much of their professional line lately.

ZFS sounds like a high-professional sort of thing, and Apple doesn't seem to care about professionals anymore. Maybe prosumers, but not professionals. Pros definitely are not the majority of Apple's consumer audience, either.

And unlike Thunderbolt, ZFS doesn't sound like something they can market and make mass profit-margins off of. Thunderbolt has a lot of dongle potential, ZFS probably doesn't. They'd probably also have to have a big hand in the development, which I just don't have much faith in Apple attempting to do.

But personally, I'd feel kind of paranoid about keeping all of my data on one gigantic harddrive... That's a lot of space to keep in one place, even if backed up somewhere else.

 

[ChrisA]

Agreed, Apple has no storage products, and none likely coming. But, as you say, Apple would point you at a Promise or G-Tech RAID box...that could be up to 64TB. At the rate of HD improvements and storage growth, we could be pushing 100TB HFS arrays in a couple years.

How would that be formatted and used by a pro user or small business owner? Right now....we know it is gonna be HFS via Disk Utility 99+% of the time. There is no other realistic option outside of the enterprise world.

Most experts agree that such large a storage array needs a better file system: more reliable, more efficient, snap shot-able, self healing, etc. ....

You are right about everything above. But those 100TB RAID boxes will be NAS boxes that use their own file system. It might even ZFS but most users would never know that. In other words, to Mac OS X, the big storage box will look like a bunch of folders and files not like a bunch of disk sectors.

If you want a ZFS based NAS connected to an Apple computer you can have it today. It is very reliable and mature. See this http://www.freenas.org/about/features.html
This is free software based on BSD. You can build your own system to buy a turn-key system.

Here is one BIG issue: Notice the hardware requirements. A ZFS based NAS does not run on a low-end processor. They are using/recommending something like a quad-core Intel chip with 16GB RAM and some high-end disk controllers and network cards. ZFS was developed by Sun to run on Sun Servers, there are $100K servers. ZFS depends on the availability of lots of cheap compute power and lots of cheap RAM. Today you don't need a $100K server box but the minimum is about $1K for an entry level built it yourself ZFS file server.

Most people building file servers use a Linux file system that is very much like HFS+ because it requires less expensive hardware.

NOW, to answer your question. ZFS will be popular when a 64-bit, quad core system with 16GB RAM and hardware RAID sells for $250.

OK, yes you CAN set up a BSD/ZFS system on smaller hardware. I've done it. Performance is poor but at least you can learn about ZFS.

 

[hobowankenobi]

Agreed. I get it. Really.

But still dancing around the real issue: What will average, normal Mac users do?

Yes, ZFS is not for casual, non-tech users. Yes, NAS is an option now. Yes, a tech user can have NAS + ZFS (or similar robust file systems) now, if they build their own.

But the question is when will Apple implement a modern file system? To be clear, I am not asking because I need it; I am asking because having worked in Mac IT and for an Apple reseller, and now at a University supporting Mac users....I can assure everyone that 90 something percent of Mac users still buy and plug in locally attached storage. Not NAS, nor do they build their own anything.

So I suggest there are only 4 paths:

1. Apple radically improves/updates/replaces HFS in a future OS release to handle large local, Mac formatted, bootable storage.
2. Apple introduces a new option that has a ZFS-like feature set specifically for locally attached storage (likely non-bootable)
3. Apple stays with HFS as-is; perhaps minor tweaks, but nothing to address 50TB+ local storage. They wait/expect third-party/NAS/Cloud/Something new to address the issue for large storage needs.
4. Apple fully abandons the SOHO/Pro user market completely.

Some folks argue they have been moving away from the Pro user model for some time. Doing nothing would reinforce this perception, and open the doors to rivals for (what has been) a lucrative high end market.

This idea of abandonment is a bit at odds with other things though. Some thought there would be no new Mac Pro, yet we have a serious Pro level work station. One could argue that Thunderbolt is a bit overkill for small locally attached storage, and the money and resources needed to create it were wasted in light of USB 3 and beyond. And there is little need of TB 2, much less TB 3, which is in the works. If they really wanted users to plug in third party "enterprise" storage, I would expect them to introduce stuff like a native iSCSI initiator, or AoE support, or some similar protocol for easily accessing data at local speeds. Or 10Gig NICs. Or something showing direction on very large, very fast storage connections.

Obviously I am swimming up stream here, but I certainly hope they are working on #1 or #2. I see small businesses using and managing Macs more than ever before, often without dedicated IT. I hope to see this growth continue, and local storage that is "just works" one of many tools that can help this trend continue.

 

[hobowankenobi]

I don't think Apple ever intends to take mass local-storage seriously. You can see it in their iPhones and iPads because, price aside, they can get away with tiny SSDs. You can also infer it from the platter drives in their iMacs. You can also infer it from how they've cut out so much of their professional line lately.

ZFS sounds like a high-professional sort of thing, and Apple doesn't seem to care about professionals anymore. Maybe prosumers, but not professionals. Pros definitely are not the majority of Apple's consumer audience, either.

And unlike Thunderbolt, ZFS doesn't sound like something they can market and make mass profit-margins off of. Thunderbolt has a lot of dongle potential, ZFS probably doesn't. They'd probably also have to have a big hand in the development, which I just don't have much faith in Apple attempting to do.

But personally, I'd feel kind of paranoid about keeping all of my data on one gigantic harddrive... That's a lot of space to keep in one place, even if backed up somewhere else.

This may be the case. Would be #3 or #4 in my list above.

Yes, the bigger the storage needs get, the worse the problem. Because of the needs for backups, it is really exponential. And while HDs are getting faster, cheaper, and bigger....they are not getting more reliable.

Moving around TBs of data take too damn long without large, fast, reliable local attached storage for most folks. Cloud and low cost NAS? Forget it. Way to slow for most users beyond near line storage, and incremental backups. Not acceptable to anybody shooting HD and RAW. Not needed by the large majority of iMac users....but still needed.

So yeah...your points are valid, but yet there is still a very real need for the foreseeable future. Perhaps you are right: they will simply abandon everything "Pro". I just hope that is not the case.

 

[Morris]

It's a myth that file systems such as ZFS can only be useful for large storage solutions.

ZFS (or BTRFS or similar) can be useful for anyone who stores valuable data or needs reliability. Even if it's just a couple of megabytes. The key is in the checksum feature of ZFS (and others) that protects your data against bitrot. Most backup situations won't protect you against bitrot (because you backup corrupted data) and even RAID won't always prevent bitrot. Your valuable photos, documents etc. could be slowly disappearing without a warning. Years later you try to load a document only to discover that it's corrupt and you backed up the corrupt file. Or what if it's not a document but a vital system file that has slowly become corrupt?

That is where using a checksum to guarantee data integrity comes in. With checksums you know that the data as it's on the storage medium is exactly like when it was written (whether seconds or years ago). Ideally it can even automatically be repaired.

I would very much like Apple to integrate a scaled down version of ZFS. Strip out the deduplication and complex volume management (which are mainly aimed at large storage solutions) but keep the checksum and snapshots (for Time Machine). The benefit would be that the code is tried and tested in serious situations (who dares to store their precious data to a brand new file system?) and many tried and tested tools (partitioning, troubleshooting, repair etc.) are already available. You could even attach an OS X ZFS drive to any UNIX machine with a full-scale ZFS implementation and read it. Because it would be stripped down you wouldn't have to deal with the high RAM demands that features such as deduplication have.

Anyway, I don't care if it's ZFS, BTRFS or some other filesystem, as long as it has checksums, extremely flexible partitioning (resize any partition on the fly, without shutdown) and is prepared for a future where 100GB files on 100TB disks are common.

 

[2984839]

I don't agree that ZFS is only for large corporate users or that it requires a screaming fast server. I used it on a laptop for years running FreeBSD and was satisfied with the performance, and that was with 4 GB of RAM and an i5-540M. Casual users will likely not use the full set of features that ZFS offers, but that's not a problem; Apple would probably hide those features behind a GUI anyway.

The real problem with is ZFS is its complexity. Not only is the list of source files huge, it is difficult to reduce functionality to get only the features you want. It wants you to implement all or nothing. Sure, it could be done, but it would be a big task. It may not be worth it over some alternatives.

 

[gnasher729]

Every single application on the Mac expects there to be a POSIX compatible file system. Lots of applications expect certain behaviour as it is implemented by HFS+. If that behaviour changes, stuff will break. So unless Apple creates a compatibility layer that maps for example HFS+ calls to ZFS calls, it won't work. And any new ZFS features will not be used, because nobody uses them.

The real question is: What's in it for Apple? How many Macs would they sell more if there was ZFS available? On the other hand, how much would Apple have to pay for licensing, development, support? Can they make more money by paying the same licensing, development and support in some other area?

Put differently: When my wife wants to buy a new Mac, and the sales person tells her "the new Mac comes with ZFS instead of HFS+", is she more likely to buy, or is she more likely to run away in fear of dangerous TLAs? I think she and many customers will more likely run away.

 

[Pakaku]

Put differently: When my wife wants to buy a new Mac, and the sales person tells her "the new Mac comes with ZFS instead of HFS+", is she more likely to buy, or is she more likely to run away in fear of dangerous TLAs? I think she and many customers will more likely run away.

ZFS doesn't sound sparkly and magical enough. It's actually Apple's Smart Storage™!

 

[Farsider]

The future of storage is either Cloud based or in a on-prem centralised form (for file, that means NAS).

NetApp have a great file system (WAFL) that's proprietary to them. Its arguable the Grandfather of ZFS. Its just too bad they don't offer a consumer option.

 

[hobowankenobi]

Every single application on the Mac expects there to be a POSIX compatible file system. Lots of applications expect certain behaviour as it is implemented by HFS+. If that behaviour changes, stuff will break. So unless Apple creates a compatibility layer that maps for example HFS+ calls to ZFS calls, it won't work. And any new ZFS features will not be used, because nobody uses them.

The real question is: What's in it for Apple? How many Macs would they sell more if there was ZFS available? On the other hand, how much would Apple have to pay for licensing, development, support? Can they make more money by paying the same licensing, development and support in some other area?

Put differently: When my wife wants to buy a new Mac, and the sales person tells her "the new Mac comes with ZFS instead of HFS+", is she more likely to buy, or is she more likely to run away in fear of dangerous TLAs? I think she and many customers will more likely run away.

I think this is mostly a false argument. Most consumers don't know or care about file systems, and that won't change. They just want their data safe, backed up, and fast.

The point of the original question is....Apple will have to eventually change. They have several times before, and there is no reason to think they won't again.

I do remember when they introduced journaling about 10.2....and that is about as unsexy as it gets. Yet it was needed to make the file system more robust, which makes the machine happy and more fault tolerant, and that makes Mac consumers happy.

Or how about when they added ACLs about 10.5 so that Macs could play nice in a mixed-platform environment? That was certainly not in any adds, and only IT monkeys knew or cared....and that is back when Apple was a serious outcast in most IT shops.

In both cases....they added/changed underlying tech so that everything "just works". That is something we should expect to continue.

HFS cannot stand (as-is) forever, as other platforms move ahead. Fast forward a decade....and Apple would be the laughing stock of the tech industry if everyone else has moved on, and they have not. Sure, cloud storage will continue to grow, but it may be a long while before consumers have NO local storage. How about the latest SSDs.....that are 15TB?

 

[hobowankenobi]

ZFS doesn't sound sparkly and magical enough. It's actually Apple's Smart Storage™!

Yep. Should they decide to actively market it as an awesome new feature, it will get a sexy name. That we all can agree on.

Hybrid drive....not sexy. Fusion Drive = sexy!

 

[rekhyt]

It's a myth that file systems such as ZFS can only be useful for large storage solutions.

ZFS (or BTRFS or similar) can be useful for anyone who stores valuable data or needs reliability. Even if it's just a couple of megabytes. The key is in the checksum feature of ZFS (and others) that protects your data against bitrot. Most backup situations won't protect you against bitrot (because you backup corrupted data) and even RAID won't always prevent bitrot. Your valuable photos, documents etc. could be slowly disappearing without a warning. Years later you try to load a document only to discover that it's corrupt and you backed up the corrupt file. Or what if it's not a document but a vital system file that has slowly become corrupt?

That is where using a checksum to guarantee data integrity comes in. With checksums you know that the data as it's on the storage medium is exactly like when it was written (whether seconds or years ago). Ideally it can even automatically be repaired.

...

That's the most compelling point for the end-user.

Don't really have anything to contribute to this thread, but it's been really interesting.

 

[wubsylol]

Most of the problems with the current iteration of HFS are totally transparent to the average Mac user.

Apple can't sell a new filesystem as a new feature without first demonstrating, in layman's terms, what was wrong with the old one and why it was not possible to fix the problem until now.

 

[Fuchal]

HFS++? HFS#? Objective-HFS?

Sorry... It's Friday and I couldn't resist being silly for a moment.

HFS Pro

 

[2984839]

Most of the problems with the current iteration of HFS are totally transparent to the average Mac user.

Apple can't sell a new filesystem as a new feature without first demonstrating, in layman's terms, what was wrong with the old one and why it was not possible to fix the problem until now.

I don't see why they would need to sell a new filesystem as a feature. They barely mentioned journaling when they added it to HFS+. There's really no way to get regular people excited about a filesystem.

 

[Erdbeertorte]

Leaving aside the features like better data integrity protection, I don't understand the mentioned problems with larger storage now and in the future when HFS+ is stated to support volumes and file sizes both up to 8 Exabyte and a maximum number of 4,294,967,295 files for each volume.

Does this just work in theory and are there much lower limits in reality?

Could anyone please explain the issues to a stupid person like me?

 

[Markoth]

Well, personally, I've been using OpenZFS on OS X for awhile now, and it's working great! Just don't use it with a USB-attached drive (including the SD Card slot). Seems there are still some unresolved issues there. All-in-all, the space I'm saving from ZFS's compression feature has been well worth it! I'm using a combination of GZIP-9 (for when performance isn't an issue) and LZ4 (which tends to actually result in a performance boost). I've gained somewhere around the area of an extra 30GB on this 128GB internal SSD, just from ZFS. I don't really care if Apple implements something better at this point. ZFS is working just fine. The only downside is you can't boot from ZFS, so the boot partition still has to be either HFS+ or Core Storage.

Which brings up another point. Core Storage IS the future of HFS+. It has all the signs of someday becoming what ZFS is today. Snapshots, RAID and compression could easily become features of Core Storage in the future. Whether it will happen depends entirely on Apple, but the foundation is definitely there.

I don't see why they would need to sell a new filesystem as a feature. They barely mentioned journaling when they added it to HFS+. There's really no way to get regular people excited about a filesystem.

Exactly. The goal is to have the user NOT notice the filesystem. When a computer malfunctions less, the user ends up being happier, but they couldn't tell you why. All they know is they're NOT having issues. That's the goal of a filesystem: To do its job well. If a filesystem can't do that (HFS+ for example), then it gets noticed.