You do understand that the 'SS' in SSD means solid state, as in DRAM? All RAM chips have parity built into them, unless of course, you specify non-parity, which I seriously doubt any SSD maker uses. That's why I mentioned that the parity would be at the data layer, as opposed to the block or stripe, as it were. In the event of a parity error on a chip, most OS/BIOS will report something like: 'Panic: R/C 54, Offset; 055769=0058CB4D3 DDRAM Parity Error'.
The other advantages that I didn't mention, but are a function of RAID 0 and SSD is the very low latency. We don't have to worry about spindle location(s) and seek/settle times compared to rotational mag media. in a RAID architecture using spinning media the latency plays a significant role when high write content data is in use. Some modern commercial disks are now using a dynamic caching algorithm on the disk chassis, but the majority of disks are read cache(naturally), so that successive writes can't destage efficiently due to the latency/seek/settle parameters, which aren't found in SSD.
Now, the one type of delay that isn't helped by using SSD is the read-modify-write penalty found in RAID 5. Of course, we weren't using 5, so that's immaterial.
RAID 0 is 'protected' at the data layer by the parity calculation once the write reaches the chip level. Of course, DDRAM, or any kind of non-volatile memory chip is susceptable to loss in the event of a power fault. One could get flash type memory and overcome that limitation, but the access times of non-volatile memory chips is many times slower than your typical DDRAM. Of course, once the bytes are assembled at the IO controller layer, getting them to the DRAM chip is subject to error, but that's a very low failure probability.
If I were designing an SSD system, I would use a hybrid of DDRAM for the controller cache to destage data from the IO controller fast, then I would use a battery system to keep alive in the event of a power fault. The main memory bay would consist of double parity non-volatile RAM(flash) type memory. This is typically what's found in modern cameras, and those little SD flash memory sticks. Dynamic R/W cache allocation, and quad, or octo 32 bit data paths to the server via a double board PCI+ IO.
