On Sun, 2009-02-08 at 15:33 -0600, Sam wrote:
The software raid in linux with mdadm is very powerful. Alot of people stay away from software raid because they think that a hardware solution would be easier to work with. But with a hardware solution, how do you monitor the status of your drives? There is usually windows software for that but normally a linux client is non existent. All of the monitoring and management is built into mdadm. Once you learn it, it is very easy to use and you can move your raid array from system to system as long as mdadm is installed. You certainly can't move a hardware raid setup to another machine unless the cards are identical.
While I think that Linux software RAID is both solid and stable, when running a production environment I'd much rather use hardware RAID with hot-swappable drives. Example? Dell PERC RAID. Yes - historically there have been problems - but today it's rock solid. Monitoring it? Easy - there are Nagios plugins for omreport. Drive fails? Pull the drive and put the new one in. Nothing else to do. Same thing with HP DL-[35]xx class boxes...
And if you're running, say, a farm of a few hundred servers, you can just have someone go in once a week armed with a list of disks to pull and replace.
In short, IMHO, hardware RAID with hot-swap capabilities, on proven, STANDARDIZED hardware makes it easier (and cheaper) to support a larger number of boxes. (If you don't have standardized hardware, and tend to run somewhat of a mishmash, you're probably better off considering software RAID...)
-I