As much as I like to put the Drobos in the same shelf as other overly expensive and mediocre performing products like those from Bose or Apple, they do have some attractive qualities.
The newer Drobos like the Drobo S offer dual redundancy like RAID 6, which I think is important. In RAID 5 and even RAID 1, any failures during the rebuild period and you can lose everything. With RAID 6/ Drobo's Dual Redundancy mode, a failure during a rebuild still affords you RAID 5-level single-disk failure protection. And the Drobo is data aware and rebuild times depend on the amount of data you have, unlike traditional RAID's need to rebuild the entire disk, which can make the rebuild quicker in theory.
And there is the issue of bit-rot, where the data on the harddrive is altered on its own. The Drobos perform a checksum on every read due to the nature of its storage virtualization system and can detect unexpected data changes. I've seen some RAID cards offer scheduling of refresh scans of the array to make sure all the redundant data matches the actual data.
How do people protect against failure during a RAID 1 rebuild? Is there a version of RAID 1 where a copy of the data is duplicated onto two other disks instead of one?
The first Drobos were a complete joke with their 10-20MB/s speed, but the Drobo S' 80MB/s speed makes it just fast enough to be not annoying. Too bad about the Drobo S' $800 price with zero drives and annoying requirement for Port Multiplier support. I suppose one can justify the price saying that a good RAID controller and drive case will cost one at least $500-$600, but they forget to say that it will perform at least 4 times faster.
Now, why the heck does hardware RAID cards cost so much? It's not as if the silicon is particularly complex and dense, especially when compared to current CPUs and GPUs. Sure it's a custom single-purpose chip, but should it cost several times more than a CPU built with 32nm 400 million transistors? I would not be surprised if those chips were designed at least 10 years ago. I'm convinced it is related to the low-volume business-oriented market. Given that, the Drobo's looks even more overpriced, since they contain general purpose CPUs and RAM to run their BeyondRAID algorithm. What makes hardware RAID cards so special that it outpaces software raid running on general purpose Multi-core 3+GHz CPU and 8+GB of ram?
I do like seeing a fault-tolerant storage system like the Drobo that is designed to 'just work' for everyone, but too bad it's priced for the elite.