[Update november 27th; DiY FusionDrive confirmed!, see last page]
[Update november 9th; Do-it-yourself FusionDrive available!, see last page]
Hi all.
For a change, a regular blog post, not a how-to or a script to share, just some info on a “new” technology “invented” by Apple; Fusion Drive.
First of all; what is it?
Well, that’s a difficult question, since only Apple really knows and there probably isn’t a Mac outside of Apple HQ that runs on it. But based on the information presented by Apple and discussed by many techfora, like arstechnica and ubergizmo, already discussed Fusion Drive in depth and here’s the scuttlebutt;
1. Apple Fusion Drive combines a regular HDD with a fast SSD so files accessed often are available at high speeds, others are on the large capacity storage. All seamless and even so that certain files of one project can be on the SSD (the files you use) and others on the HDD (the documentation). It’s file-based (actually block-based), it’s seamless, it’s automatic. It fits the Apple philosophy perfectly; it’s dead-easy to use.
2. Apple Fusion Drive is NOT a cache; A cache is typically a volatile data storage with very high speed. Caching has been around for ages; VGA-ROM and BIOS were cached in RAM as early as the 90’s. CPU’s have L1 and L2 cache, L1 being insanely fast, L2 slower but still faster than most RAM. Your HDD has a cache; it’s usually about 32 or 64 MB, not much, but it’s there. High-traffic websites do caching as well, and although different in nature, the effect is the same; Varnish, for example, is a cache for Apache; the original website still is there underneath, but you get a fast copy served instead.
Common denominator of all these is this; The data exists in both the original location (be it a CPU register, a block of Data in RAM or on your Harddisk) AND the cache. That’s what caching is; a far far faster accessible copy of your data. If the cache is destroyed? nothing to worry about; just get the data from the slower medium and you’re done.
3. Apple Fusion Drive does not take away capacity (actually, it does, about 4 GB is reserved, but the big picture is; no; you don’t loose capacity, not in the way that is “lost” with caching); With Apple Fusion Drive, the capacity of both the HDD and the SSD (minus 4 GB or so) is combined to form one big logical volume. If you have a 1 TB HDD and a 128 GB SSD, you will end up with a 1.12 TB Fusion Drive. Of course, capacity isn’t lost with caching, but when adding a 128 GB SSD to a 1 TB HDD and still ending up with 1 TB of storage, well, you could say, you lost 128 GB of capacity, right?
This technology used by Apple Fusion Drive is called Data Tiering.
(read on…)
Well, this weekend i achieved a working Fusion drive on my “old” Imac (late 2009). I replaced the internal HD with a larger disk, replaced the optical drive with a 128GB SSD.
Did the software configuration, installed TRIM support (needed for aftermarket SSD’s) and my bootup time is now 20 seconds !!!!!
Fusion drive rocks ! (and can work in older imacs)
All frequently used apps start very fast.
Excellent! Haven’t (still!) tried it myself. Did it go as smoothly as the video claimed? (See also last part of this post)
The software config was the most easy part. This went perfect like in the video. The hardware part is the most difficult in an Imac.
After a week working with a Fusion Drive, everything else is slow :)
After the official warranty period i will do the same upgrade to my macbook.
In your post you mention that Fusion Drive is file-based. In fact, all currently available data indicates that it is actually a block-based technology, only copying the most used data on a block by block basis. This theoretically has significant ramifications for performance, security, and reliability.
You are absolutely right about FD being a block-based technology, but I think you are wrong about the ramifications for performance, security and reliability.
Reliability: this is no more reliable or unreliable than a Raid 0 (striping) setup.
Security: the security of any filesystem is OS based, this is no different. I can choose to do full-system-encryption on a FD so a complete set of a FD-“array” is secure, and as any striped volume setup, without the other, the one is useless. Sure you can get bits and pieces, but that’s always the case.
Finally: performance: Well, no SSD part is slower than some SSD part and if you need ALL of the file often, then ALL of the file will be on the SSD. No problem there.