Taz said:
Thanks for the feedback, the peformance gain from raid 0 is in a great deal?
To clarify what David said, striping writes different pieces of a file to different disks, so
both disks in the set have to be functional, or have additional redundant copies to be able to read your data back.
The performance gain is not significant for normal consumer purposes, IMHO
. The rest of your system has to have the bandwidth to take advantage of the striping, and it's mainly useful for high-demand data like video editting, enterprise databases, etc. In those cases, the preferred interface is SCSI instead of ATA, although eSATA is starting to get there. Add to this that many consumer-grade RAIDs are implemented via software on your computer, and you have another potential bottleneck if you are looking mainly for a performance boost.
Ideally, each disk in the array is also running on its own drive controller.
Striping was particularly useful back when drives had minimal to no on-board buffering of data, and the drive platters and head seeks were significantly slower than those of modern drives. There is a whole bunch of CompSci theory around how best to organize data on a disk to minimize latency, head seeks, and other stuff. In the old days, having more "spindles", aka drive mechanisms, tended to improve those efficiencies.
FWIW, you can also run RAID 0+1, 10, or 50, which gives you striping and redundancy, but uses even more drive mechanisms. (They all have slightly different performance and redundancy characteristics).
RAID=Redundant Array of Inexpensive Disks
Here's a link to the original Berkeley paper from 1987 describing the concept of RAIDs:
http://digitalassets.lib.berkeley.edu/techreports/ucb/text/CSD-87-391.pdf
As mentioned previously, RAID is NOT a backup. It supplies redundancy for certain types of hardware failures. If you get hacked, or accidentally nuke your data, it's gone. RAID is particularly useful for people that can't afford downtime due to a hardware failure, even just the time it takes to reload a backup.
Offline backups are the only true form of backup, whether that's a simple snapshot of a disk image to another drive, a formal incremental file backup, or spooling off to tape or other media. To be really safe, you should occassionally reload one of your offline backups to an active volume to insure 1) that it actually got writtten out properly the first time *, and 2) that the backup media is still intact.
* I know of one company that had a really consistent backup plan, with weekly full backups, and nightly incremental backups. Unfortunately, they discovered the hard way that they were writing only the
directories of their files to the backup volumes, and were not including the files themselves... A cursory look at the backup log would show that a whole series of directories had been traversed, so at first glance, everything appeared correct...