Strange raid question

[prelude2250]

Hey strange question. Companies have been paralleling things to make them better for a while now. i.e. dual & quad core cpu's, 4870x2/gtx295,etc. My question is why dont they do that with hard drives? I understand that there is raid but its not in the same unit. (I also understand that there are raid boxes) I also know that they have produced hd's with dual read/write heads a long time ago. Why cant they take 2 7200rpm sata laptop drives and fit them in a desktop hard drive case with raid and hd controller-board and have it set to run raid automatically from the factory? It seems like a good idea. What am i missing? I also have and would continue to have the traditional raid setup but there would be situations where packing that in one hard drive could be useful.

 

[sminlal]

Well for one thing putting two drives in a standard 3-1/2" form factor drive would mean that the disks could be at most about 2" or maybe 2.5" in diameter. You'd end up with less available space, even if you used RAID 0 (which wouldn't be a great idea because of the poorer reliability). You'd probably also have to use separated chambers so that the two rotating stacks of platters wouldn't screw up the airflow.

Another concern is that very high performance drives already have to deal with vibration issues when multiple drives in an enclosure affect one another - two actuators in the same drive case would make that even more of problem.

Parallelism in chips where you basically have no better way to spend your transistor budget these days is one thing - but doing it in very mechanical devices like a hard drive is something else altogether.

 

[sub mesa]

SSDs use interleaving (RAID0) internally, for example the Intel X25-M has 8 parallel channels, or a virtual RAID-array of 8 drives.

The major advantage of SSDs is that they can work fully parallel; whereas a HDD can only do one thing at once since there is only one head. The difference ofcourse is that an SSD can perform up to 35.000 IOps and a HDD can be as low as 100-200 IOps where it is bottlenecked by the seek rate of the head.

So, at least in theory, an SSD could read from all its chips at the same time, and get virtually unlimited speed (no reason an SSD can't do 1GB/s or more). There is virtually unlimited possibility to go faster, because there is no single component (like a head) which everything has to wait for. All chips can work independently and alot of data can be processed at the same time in parallel.