Installation (continued):
The next screen shot shows that once formatted and mounted, the resulting folder size for a four drive striped array created with LVM corresponds to what would be expected when using four 512MB drives in such an arrangement.
I ran a few tests of arrays configured using the LVM, and the results were a mixed bag. The reliability of the arrays was the main problem, as sometimes I could read/write only a few times before the array 'died', and other times I would get I/O read/write errors on my first attempt to access the drives. The performance was as expected (when it worked), but the limited number of configurations possible and the unreliable performance was less than impressive.
I had actually tried mdadm first, but gave up on it for the simple graphical interface of LVM before conducting too many tests. mdadm has a command line interface, and requires more typing than clicking in order to get your arrays configured. Not difficult by any means, and once the LVM approach proved to be too limited for my interests, I gave it another shot.
The screen shot below take a look at some of the steps required in order to set up a thumb drive RAID array using mdadm.
Once comfortable with the commands required, mdadm allows many configurations of RAID, including; RAID 0, RAID 1, RAID 5, and Linear. I began experimenting with mdadm, and found a similar issue with array reliability. I can't say for sure what the problem was, as it could be related to one of many factors, including; operator error (or ignorance), software configuration, software limitations, hardware quality, hardware compatibility, or perhaps a saturation of the USB bus.
After the sporadic functionality in Linux, I connected two of the drives to a Windows XP machine again just to see if they performed as poorly. Although no RAID arrays were created in Windows, I wanted to see if normal operation resulted in the same types of failures. Sadly, the drives worked just as poorly in Windows... Sometimes they wouldn't power on, sometimes they gave Windows errors as being unrecognized, and some times they would just quit in the middle of the transfer. This made me feel marginally better about my understanding of what I was trying to do in Linux, but was disappointing to think of how much time I wasted trying to work around these inferior drives.
The reliability issues slowed the testing process down greatly, but eventually I was able to complete enough runs in all of the arrays I found to be of interest. The next section will take a look at the results acheived using mdadm only.
