To get the best performance out of multiple drives you want to try to balance the I/O load across both of them as evenly as possible. In theory, if each drive is doing half of the total work then you'll get optimum performance.
RAID 0 does this by alternating I/O operations between two drives based on their logical block number. Depending on the stripe size, when you write data to RAID 0 sequentially it writes "n" blocks to one drive, then "n" blocks to the other drive, and so on.
But here's the rub: there are two different types of disk performance - sequential transfer rates and random access times. RAID-0 splitting the I/O load across two drives works really well for sequential I/O - but it can struggle just as badly as individual disks do under a random I/O load.
For example, if you're making a copy of a large file, the heads are going to have to move back and forth between the old and new files. That can cause head "thrashing" and reduce performance, even on RAID-0. But with two independent drives if you can arrange for the old file to be on one drive and the new file to be on the other drive than the copy can proceed without any head thrashing at all, and it can be much faster than you'd get on a RAID-0 volume.
So you can't just assume that RAID-0 is a slam-dunk to solve every performance issue. You have to choose your disk organization with a good understanding of what your workload is. If you're able to identify which files are active and put them on separate drives, there's a good chance you'll be better off than if they were part of a RAID-0 set.
Message edited by sminlal on 10-02-2011 at 11:18:55 AM