Is the Drobo 'storage robot' fact or myth?

[UPDATE - See updated review for a much more in-depth and accurate assessment of the Drobo] There's a new personal storage device that's been getting some rave reviews lately.  The company is Data Robotics and they sell a device called the "Drobo" which is a USB 2.0 connected device that houses 4 hot-swappable SATA hard drives.  Many online gadget reviewers, colleagues, and friends of mine were wowed by the company's promotional video and raved about it so much that I couldn't ignore it.  One online reviewer even claimed that you can keep listening to music even when you're down to one disk, but this would seem to imply that there is 4-way mirroring (read 1/4 capacity) or a lot of luck that the music you're listening to happens to reside on the sole remaining drive.

All this talk got my curiosity going and it put me in my "myth busters" mode so I decided to look in to these fantastic claims.  I carefully reviewed the video demo which showed a man adding a third hard drive while watching a short clip of the movie 300 and then yanking out a different drive with no loss in data.  The person then pulls out one hard drive and then lets the Drobo repair itself in to a healthy state in a matter of seconds and then pulls out a second drive leaving only one drive in while watching the movie clip.

The person giving the demonstration mentions that the redundancy mechanism works intelligently at the file level so that it doesn't need to replicate the entire drive.  That's great when you only have a tiny portion of the device in use but it's just as slow as traditional RAID when you have the drive filled near capacity.  Another neat trick is that the drive seemed to have gone from some sort of file-level parity-based redundancy mode in to a mirroring-based redundancy mode (we can think of this as the equivalent of going from RAID5 to RAID1 transparently).  But this neat trick (going from 3 to 1 drive) wouldn't work at all if you had more data than the smallest drive in the array so is this really useful in reality?  Now I must admit this is some fancy software engineering how they managed to make this transition so seamless to the user, but I need to warn my readers that this doesn't work in the real world because it's very unlikely you're going to have so little data in such an expensive external storage device.

The other major selling point of the Drobo is the "pay as you grow" feature.  Data Robotics even claims on their products page that:

Hard drives get bigger and cheaper all the time. Don't buy storage capacity until you need it. Buy capacity "just-in-time" possibly saving you hundreds of dollars.

Oh really?  It's true that hard drives get cheaper over time, but can you really mix drive sizes in a redundant array?  That certainly seems to be what Data Robotics is implying with their "pay as you grow" motto but I dug deeper in to their capacity calculator to see what's really going on.  I took a Drobo system that had 3 100-GB hard drives which had an effective capacity of 200 GBs and decided to add a 1-TB (TB = 1000 GB in simple terms) to the array.  I looked at the effective capacity gained and it was a measly 100 GBs gained (see screen capture below taken from Data Robotics' capacity calculator).

Note: The screen capture below shows less than 100 GB gained because they're using the more conservative binary math method of computation.  I'm using the simpler to understand decimal math calculations which I feel are more useful for human consumption.

So wait a minute; you add a 1000 GB drive to your Drobo box and you get an extra 100 GBs of capacity!  Their motto should really be "pay-through-the-nose as you grow"! [Update 5/30/2007 - There are better scenarios that does allow better use of mixed drive sizes. See updated review] This doesn't make sense at all once you understand that there is no free lunch and you can't violate the laws of storage redundancy.  You'd be much better off if you simply purchased a dirt-cheap 100 GB drive because it would have increased your capacity by the exact same amount.  A much better way to go is to migrate to four cheap 300 GB drives which gets you 1200 GBs of effective capacity.  You'd also have to toss the old 100 GB drives which goes against the spirit of "pay as you grow".

The bottom line is that Data Robotics managed to put in some slick and automated fault tolerance features that allows it to expand in a more efficient manner and it also has the ability to drop capacity to maintain redundancy in the situation where multiple drives may fail.  The problem is that this only works when there is very little data in the unit and it pretty much works like any other RAID device when the device is mostly full.  The "pay as you grow" feature is highly questionable since it doesn't make any sense to add larger capacity drives than the ones already in the unit and you can do the same kind of hot-expand with traditional RAID devices, just not automatically.

From a performance stand point you have a device that's limited to the speed of the USB 2.0 port which is typically somewhere between 10 and 30 MB/sec (80 to 240 mbps) so it's nearly 10 times slower than 2.5 TB server (2 TB effective storage in redundant mode) I built on the cheap.  The USB interface also means that the Drobo cannot be shared by multiple computers on a network (without a PC acting as a server) and it can only be attached to one computer at a time.  I will admit that the device looks fairly solid mechanically and it looks like it's very easy to use.  This may be perfectly suitable for many people, but for $700 [Update - Price has dropped to $500] (no drives or PC included) I'm not impressed by this relatively slow and off-network storage device even with the reduced price.  Be sure to check out this previous blog "Solving the network storage dilemma" for a better understanding on various storage options.