Interestingly enough, I just had a 2TB drive go bad in my IBM EXP3000 12 drive DAS box. For a rough comparison on the topic of real RAID vs fake RAID, I had a 5x 1.5TB RAID 5 array (7.5TB total, 6TB usable) on an Intel 5000X ESB2 chipset fake-RAID that also had a drive go bad a few years ago. The rebuild process took 36 hours to complete, which is a scary-long time when you can only suffer a single drive failure in RAID 5 before you lose everything.

This new system is using an Adaptec 6445 PCIe HBA RAID card with 12x 2TB drives configured in RAID 6 (more parity, can lose up to 2 drives, so less worry about a failure). It only took 6.5 hours to rebuild the entire 24TB RAID array. That's 3.2 times the capacity at 5.5 times faster the rebuild speed.

Oh, and as to why I chose RAID 5 and 6 instead of RAID 10, for my use I wanted maximum capacity and fault tolerance, and write speed was relatively unimportant for my purpose. This array stores all my digital movies, so I only need write speed when first moving the files over, then it's all about the read speed.

Thanks Ozzfan, that's exactly the info I've been looking for, as I have 12 drives (combo of 1/1.5/2tb) currently in the red :-(

i just bought a Asus R.O.G raidr Pci-E SSD 500gb i plan to use as primary...looking forward to picking it up later tonight...200euro :)

had 2 samsung 850 pro 256's in raid0 ...sold them both today for 110 each ..made a 20euro profit and makes things a lot easier (cables)

had 2 samsung 850 pro 256's in raid0

That's exactly what I use in my daily driver. Bought each one for $250 roughly 3 years ago. The speed is amazing!

Charlie, a question you may be able to answer me, experience and all.

What is better to run on 3 drives:
- RAID1 on two drives plus a hot spare?
- RAID5 (although slow and archaic)?
- RAID10 (not RAID1+0)?

I now have my NAS set up for RAID1 plus hot spare, but because we're in a legal dispute with the store we bought the old NAS from about who needs to replace it and the damaged drives, I don't see them coming up with two replacement drives so quickly. (They're trying the silent treatment, ignoring all our letters, phone calls and emails, hoping we go away).

I have been reading around and seen that perhaps best is to use either RAID1 plus hot spare, or RAID10. But then others say "3" is a no no for anything other than mirroring plus hot spare. So, perhaps that you (or others here) know?

Thanks.

---

The answer would depend on how I would prioritize these three criteria (and this is true when trying to decide RAID level for any situation):

Storage Capacity
Fault Tolerance
Write performance

With 3 drives, if fault tolerance were top priority, go RAID 1 with hot spare (a.k.a. RAID 1E). If capacity were top priority, go RAID 5, but I'd strongly recommend a RAID card that can handle the XOR calculations. If the NAS box can't do XOR, I think RAID 1 with hot spare would be your best option.

True RAID 10, which is the same as RAID 1+0 and RAID 0+1, requires a minimum of 4 drives. This configuration would give you the best balance of write performance and fault tolerance at the cost of capacity.

Okay thanks. I'll leave it at RAID1 then for the moment, until I can get 4 drives in there.

---

A couple of notes from comments in the thread...

If you have three disks intended for RAID1 (mirror), then you should have the options for:

Use 2 active plus 1 hot spare. That gives x1 read speed and up to x2 read speed, x1/2 up to x1 write speed. You suffer an increased risk of total failure during sync to bring your hot spare online after a single disk fail, especially so if your drives are all the same age...

Use all 3 active. That gives x1 up to x3 read speed, x1/3 up to x1 write speed. You reduce your risk when syncing to a new disk after a fail due to still keeping your data live on two good disks and due to only suffering one live disk 'maxed out' for the sync...

And note that RAID10 is very different to RAID01... See hybrid RAID.


And... You can mix SSDs with slower spinning disks and in (Linux) mdadm set the slower disks to be 'write mostly' so that you preferentially take advantage of the faster read accesses to the faster drive/drives yet still take advantage of RAID with less costly devices.


Hope of interest...

Happy fast crunchin,
Martin

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)