RAID using solid state devices? Errr... why???
One of the main thrusts behind RAID is the use of multiple disk spindles to speed access to blocks on the disk. If you're using solid state devices like the Ipod shuffle or GP32 smartmedia then you don't have any seek times or latency waiting for the disk head to move to the correct sector - the bus and speed of the memory interface is always going to be the limiting factor.
As for RAID 'being useless' well, I think you're quite a way off there.
RAID is used for a blend of performance and redundancy; different RAID levels have varying amounts of each...
RAID 0 - Often called striping: Not really a true raid level as it is not redundant (the 'R' in RAID!). Data is written between two or more devices. Suppose you have 4 disks... each time you need to read or write a block of data to/from disk, the request is split into 4, with each request for 1/4 of the entire data.
This is the highest performing setup, however you can see that if any one device fails then the entire array is becomes unuseable. Read speed is up to N * drive speed, where N = number of devices in the array. Write speed can also be up to N * drive speed, but is often lower.
Capacity = N * drive capacity.
RAID 1 - Striping: Data is written precisely the same to each drive in the array.
Benefits: Data can still be written and read, even if only a single drive remains. Read speeds are often slightly increased (there is a statistically higher chance of one drive in the array already having its read head near the required sector). Write speed = Speed of a single drive.
Capacity = single drive.
RAID 3 - Striping with Parity: Requires minimum of 3 drives. Two drives are designated as a stripe set and have all the benefits of RAID0. One drive is designated as a parity drive and can be used to construct the contents of any one other drive in the stripe set if it is lost or damaged.
Benefits: Speed of RAID 0. Can survive losing one drive from the stripe set.
Drawbacks: Speed of the parity drive is a major bottleneck.
Speed = slightly lower than RAID0 due to bottleneck of constant IO to parity drive.
Capcity = (N - 1) * drive capacity
RAID 5 - Striping with rotating Parity: Requires minimum of 3 drives.
Each drive is both a stripe set and a parity disk. As the reads and writes happen to the array, a parity value is calculated and written to a drive, when the next block is written another parity value is calculated and written to the NEXT drive in the array... so each drive ends up with an equal amount of parity information... the clever bit is that from this collection of both parity info AND data on each drive, no matter which drive is lost or damaged, its contents can be restructed from the remaining members of the array.
Benefits: High read speeds - almost RAID0 levels. Can survice losing any single member of array. No single drive is a bottleneck.
Speed: Very high read speeds. Write speeds not as fast due to complicated parity checksum that needs to be computed and written.
Capacity = (N - 1) * drive capacity
RAID10 - Striping over Mirror Sets
Minimum of 4 disks. This array is constructed from two or more RAID1 sub-arrays. Data is mirrored on all disks in the RAID1 sub-volumes and then striped between the two or more mirror sets. Excellent choice for high-end use!
Benefits: very very fast IO speeds for both reading and writing.
Drawbacks: N = number of disks in eahc mirror set, Y = number of mirror sets
Can lose all but 1 disk in each of the Y mirror sets.
Capacity = Y * drive capacity
RAID0+1 - Mirrored stripe sets
Minimum of 4 disks. Similar to above, but the sub volumes are stripe sets. Again extremely high performance.
This time however the performance is paid for by the fact that if any disk is lost in a stripe set, then that entire stripe set is lost (as would a normal RAID0 be), and the array becomes in effect, a single unmirrored RAID0 array.
Drawbacks: N = number of disks in eahc stripe set, Y = number of stripe sets
Can lose all but 1 of the Y mirror sets.
Capacity = N * drive capacity
Levels 5, 10 and 0+1 and not normally found on PC's unless you have bought a very fancy hardware controller. 0 and 1 are the most common and sometimes 0+1, but are almost always performed in software on PC's.
There is also RAID50, 6, and a few others, but they are way beyond what you need to know here.
On a personal note, I've got a 4 channel IDE card in a machine at home and I run linux software RAID5 on 4 x 200Gb drives with years worth of work, documents, music etc.. stored on it. It has saved me several times in the last few years when hard drives have gone flaky or spontaneously died on me: take out bad drive, pop in new drive and restor drive from parity info... et voila!
It's not a replacement for backup, but where am i going to find a 600Gb backup device this side of 2500UKP (something like an HP Ultrium auto tape loader... at home?? LOL!)
Yes, I'm a Unix techy and admin Solaris, Linux and big Veritas disk arrays
