I need some recommendations.
Have two separate storage arrays to work with, each holds 12 sata drives.
I only have WD 1 TB drives to work with.

one array is my backup array, currently configuring it with 11 drives all as separate volume groups (technically RAID0 on each drive in storage array). Then in linux (SLES 11.3) i select all 11 drives and create an LVM with striping set to 11. So my backup volume shows up at 10.00 TB. generally if I have a problem with the backup storage array no big deal, wipe it and recreate it as it does an incremental backup every night. I chose the separate volume method to get the fastest write access so the backup happens quick.

for my data array, i am going to fully populate it with 12 1-TB drives. I don't care to have a hot spare. What's the best RAID level to choose based on what i have to work with? I have RAID 0, 1, 3, 5, and 6 to choose from. The storage array is older and connects to the system via SAS cable, but i know in the storage array software it says maximum data rate is 3 gbps. I will be formatting the file system as EXT3 in the end. In the past i've always just used RAID5 with 500gb and 750gb drives and it's been great actually, never had a real problem other than the occasional failed drive, replace it and let it rebuild overnight. wondering if there's a better RAID choice than 5 that will give me better read/write performance but still let me get a data volume at 10 TB. thanks.

> 1 TB drives

Do you care about the data? RAID 10 or 6. 5 is a bad choice with large drives because the chance of an unrecoverable read error while rebuilding -- meaning total data loss -- is high.
Do you care about how long it might take to replace a drive when one fails at 2 AM on a Saturday when you've just flown across the country? Yes? Use a hot spare.
If you don't care about the data, just stripe them for 12TB.

Of your listed choices:

0: Striping - will give you 12TB with zero redundancy. Lose any drive and you lose the entire array
1: Mirroring - will give you a single 1TB array with 12x redundancy
3: Never used it
5: Stripe + Parity - will give you a single 11TB array with 1 drive of redundancy
6: Stripe + Double Parity - will give you a single 10TB array with 2 drives of redundancy

6 will be slower than 5 due to the double parity overhead, but as AlucardZero said, with that many drives of that size, 5 will take a long time to rebuild, during which time you're susceptible to data loss. My choice would be 6 if you care about the data.

1 members found this post helpful.