RAID 0,1,and servers

[murdok5]

Okay, here is a description of what I am trying to do:

Have one namespace (share) for a group of ppl to access and store many files of varying sizes (20k < - > 10gb) with data redundancy.

Here is what I have:

3 RAID1 arrays of SCSI320 130gb baracudas, server03.



Here is what i have it setup as:

in teh raid controller, 3 raid one arrays (and a seperate raid1 for OS, thats fine). I then setup the 3 drives in windows as a spanned array, so that there is one name space, but still have raid1 redundancy at the hardware level.

The raid contoller doesnt support 0+1 (90% sure).

I realize that with multiple ppl accessing the files at once, it will have slower transfer then with seperate arrays, but that is okay. Usually there is only one person at a time.

Is there any other (more efficient/elegant) way to do this? other than adding the raid0 software level through the OS?

Mike

[CGR]

Why cant you just set it all up with Raid 5?

[murdok5]

we have tried raid5 in the past, and was not pleased (and it failed more than once). Also, with this, you can loose up to half the HDDs, and still have data, but that if only the right ones fail, which never happens.

[CGR]

When you say the Raid5 failed, how did it fail?

[murdok5]

wasnt sure. We think it was the controller, but the disks would randomlly "fail". We switched back to raid1 with same drives and it works great.

I have been doin some file tests, and it seems i setup the 3 arrays as "spanned" which tells me it will fill the first, then when you copy more move to the second, and then move tot he third as the arrays fill.

Is this correct?

[Snugglebear]

Data written to spanned drives can wind up just about anywhere on the devices. It's just an addressing abstraction consisting of one big lump of address space that the OS/filesystem will use anyway it sees fit. Files and/or parts of files can end up anywhere accross the entire span - there is no guarantee of sequential use of the physical drives. Spanning & JBOD are very dangerous, as one failure of a component drive takes out the whole volume.

Honestly, I'd look into why the controller isn't happy in a RAID5 configuration. BIOS updates usually take care of the problems, but you could also have some strange compatibility issue in there with the controller, cables, and drives. For curiosity's sake, what controller are you using?