Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
OWC blog post about 10.8.3 "roll your own" fusion drive
- Thread starter lssmit02
- Start date
- Sort by reaction score
I don't understand, the article states that earlier guides posted on the internet for creating a DIY Fusion drive (for 10.8.2) weren't actually creating a Fusion drive, but simply a CoreStorage volume.
However, the commands listed in the article are exactly the same as one of the "older" guides:
http://jollyjinx.tumblr.com/post/34638496292/fusion-drive-on-older-macs-yes-since-apple-has
This guy ran quite a bit of tests to prove that the Fusion Drive is working in 10.8.2.
What am I missing?
Edit: The article makes it sound like some change in 10.8.3 activates the tiering mechanism on the software side. Assuming OWC knows what they're talking about, does that mean a DIY "Fusion Drive" created in 10.8.2 will automatically start using the tiering mechanism when upgraded to 10.8.3, making it a "true" Fusion Drive?
However, the commands listed in the article are exactly the same as one of the "older" guides:
http://jollyjinx.tumblr.com/post/34638496292/fusion-drive-on-older-macs-yes-since-apple-has
This guy ran quite a bit of tests to prove that the Fusion Drive is working in 10.8.2.
What am I missing?
Edit: The article makes it sound like some change in 10.8.3 activates the tiering mechanism on the software side. Assuming OWC knows what they're talking about, does that mean a DIY "Fusion Drive" created in 10.8.2 will automatically start using the tiering mechanism when upgraded to 10.8.3, making it a "true" Fusion Drive?
Last edited:
There appear to be 2 primary benefits. First, after you set up your fusion drive, you do not have to worry about where the data is kept. The background corestorage features of OS X will track how you use your data, and manage it for you. Second, because of the automatic tiering, any data that you access regularly that would have been stuck on the slower, mechanical drive, will be moved to the SSD, speeding up access time to that data. For example, any photos that you regularly use that would be located on the slower mechanical drive under your setup, would now be moved to the SSD, speeding up your access to them.
Jollyjinx says using a "plain' 10.8.2 build but not sure if that worked across multiple Macs. That may be were the gap is. OWC testing across a wide variety of models and this user just doing one.
I suspect that there is a closer match between the command line tools and the GUI disk utility when used on a broader set of Macs (i.e., includes a larger number of )
Additionally, some of the arm flapping and ranting on Fusion I've seen revolves around creates one of these command line CoreStorage constructs and the GUI version getting confused when there is a problem.
I wouldn't count on that. A simple volume concatenation mapping metadata need not include the additional data needed to track block mapping candidates. The basic CoreStorage commands were present back in Lion, 10.7, but not really shed much light on till 10.8.2 special editions. if it was "all working corectly" before they probably would have introduced it earlier.
This probably needs to be constructed from creation to allow for dynamic mapping of blocks.
----------
Primarily because it works at the block, not file level. If using different volumes to segregate files then that solution only works on the file level. This solution isn't limited by that coarse grain granularity. It is very similar to the same reasons the OS manages memory layout on an extremely broad range of workloads far better than a user could by hand.
Most of the "analysis' I've seen about user managed segregation are more focused on corner cases and ignore "manual movement" time costs than are focused on common everday workloads.
Would a do-it-yourself approach to fusion have any implications on my time machine backups?
That sentence is hilarious when taken out of context
That sentence is hilarious when taken out of context
Welp, off I go starting my DIY Fusion Drive again from the 10.8.3 installer. Thankfully I have a bootable clone to restore from.
-----
It appears as a single drive and backs up to time machine without issue from my experience.
-----
It appears as a single drive and backs up to time machine without issue from my experience.
Here my question is if the SSD fails of what ever reason do I still have access to the platter data.
I just don't do this if it goes into my home folder then it's on the platter, if it goes into apps it's on the SSD..
Yup, that's the jist of it. And this eliminates the need to have huge expensive SSDs in your system as well and in the case of actual hybrid drives from Seagate and later WDC, eliminating the waste that small SSDs incur by using up a full SATA port for a measly 128GB or 256GB.
You can kinda simulate this yourself with links but what a frigging headache those are - both to setup originally and later when ya wanna reconfigure.
Seagate states that 8GB of NAND is all that's needed for a 1.08TB 7200 single platter drive to function about the same as if it were a 1TB SSD - all things considered. I personally would hedge that to 32GB or 64GB myself but whatever... As it is I have a little over 2,200,000 files on the System drive that are 4K or less in size and all total they come to less than 5GB so maybe Seagate is right? If I profile to 32K file-sizes it goes up to about 15GB tho. Of course not all of those are used very often thank gawd, but I would think slightly larger than 8GB would be more beneficial overall.
In any case the CoreStorage Fusion setups will allow me to blend however I choose. 8GB+1TB or 64GB+3TB or even 128GB+1TB if I chose to. So if this becomes popular and I hope it will, we ought to be seeing quite a few threads profiling the speeds from different configurations - which will be nice! We're going to need different tools to profile the speeds tho. The current throughput and burst-speed benchmarkers aren't really appropriate for these kinds of tests.
Last edited:
I think Mr. Fusion was exactly what was used as a backup to the time machine energy source.
If you have a hard drive with two platters, and one platter crashes, do you still have access to the data on the second platter? No. Same with Fusion. That's what backups are for.
Neither situation , tiered or not, removes the needs for backups. If talking shorter restore time there is a gap. But on the issue of loss of data, there is no huge gap.
VLIW Itantium was suppose to outclass the x86 arch. It didn't in part because in a mixed environment dynamically scheduled instructions is a winner over 100% static schedule. Sometimes at run times things are different than when looking at things static state.
For example most applications have application specific files they store in the user's Library folder. if the App is accessing those frequently then it makes since to move those also to the SSD. Ok so folks cross map and back the user directory so that most of "home" is on platter and then the library is back on the SSD. This involves sticking nose under the covers for a variety of apps.
Sure stuff like iTunes can be statically targeted with minimal dynamic losses since builky, streamed data. However, not all apps just stream data nor have parameters to play working file location.
Thats, the issue..I think I'll pass on the unified volume..
----------
I won't argue this but what I will say after moving to a new computer in the last couple days it was rather nice to fresh install ML and tie it my Lion home folder without hiccups.
Well, sure. You can pass. No one is forcing you to use a good thing. Whether the criteria you're basing your decision on is logical or not is another question tho.
A "fusion" volume as the source of a Time Machine (TM) backup? No, generally.
A 'fusion" volume as the target of a TM backup ? Yes. probably not a good idea.
You could probably 'confuse' fusion somewhat by doing daily doing full copy backups. But TM is also somewhat helped since most of the "hot access" disk metadata is going to be kept on the SSD because it is accessed so often. TM doesn't need files if they haven't changed if just doing an incremental backup.
Well, sure. You can pass. No one is forcing you to use a good thing. Whether the criteria you're basing your decision on is logical or not is another question tho.
LOL..I guess I'm a control freak. This is not the first time I've been chastised for that.
Doubtful that Apple would be so stupid as to allow TimeMachine reads to register as user and other system reads. But you often speculate to the fringes of reason so - whatever.
In either direction Fusion volumes should be able to be used (both source and destination of TM) but we don't know for sure yet - so either believe someone's speculation (guesses) or wait for the official word on it.
In either direction Fusion volumes should be able to be used (both source and destination of TM) but we don't know for sure yet - so either believe someone's speculation (guesses) or wait for the official word on it.
I really think that OWC article takes an unnecessarily conservative view of this process. I researched this as thoroughly as I could a few months ago and then installed a DIY Fusion Drive in my 2011 Mac Mini. It absolutely flies, and has been working flawlessly for months now. After asking around on AskDifferent, I ran some of the command line tests that show you what data is being written to which drive, and did find that data does seem to be moving appropriately between the HDD and SSD in my system. If you read through the comments on the OWC article, you'll see a number of people calling them out, and them responding with a bunch of stuff about the OS X terms of use — so I have a hunch they're just trying to stay on Apple's good side by not recommending people perform this upgrade on "unsupported" machines. As far as I can tell, the "special" version of Disk Utility has nothing to do with the terminal commands you issue to create the Fusion drive. That all being said, the OWC instructions were the clearest I found for the terminal commands, and I used those with zero problems. The disk upgrade hardware kit and instructions for the Mini are also top-notch.
I'd encourage you to check out posts on Ask Different, though, as a number of people have done some very deep digging into this.
Again, my experience has been overwhelmingly positive. Huge speed bump and zero issues. Time Machine and everything else works just as expected. This is such a low-level disk operation that OS X and even utilities like Carbon Copy Cloner see the two devices in a Fusion Drive as one device.
One caveat: I've seen some posts here on OS X Rumors forums about people with specific MacBook Pro models having trouble. I'd encourage anyone to spend a bit of time researching before taking the plunge.
On a more serious note....They used the whole discs for the fusion drive which eliminates the RP. This is fine if your mac supports internet recovery. If it doesn't (like a Mac Pro), then it is a bit different.
Thanks...just to be safe I rebuilt my DIY Fusion Drive using the 10.8.3 installer and restoring OSX as we speak. I had a bootable clone from CCC so it should be pretty painless (I hope).
Well, I hope your experience is like mine literally the most dramatic upgrade I've ever done. Wrestling the SSD into the Mini was a drag, but the software part was painless. Good luck!
I just came across this blog article by lloyd chambers of Macperformanceguide as he tested Fusion drives: http://macperformanceguide.com/fusion-complexity.html
Though this article came out before the release of ML 10.8.3.
He recommends having a separate physical backup of your fusion drive.
So you'll make sure that you use a single platter hard drive as well?
Everyone recommends having a separate physical backup of a non-fusion drive as well.