Recommended RAID setup?

[probedb]

Hi all,

Currently I run a RAID 5 setup on an 8 port HighPoint RocketRaid 2720SGL (x8 PCIe) card. It stands at 5 x 2TB drives giving me 8TB.

Now with larger drives available I don't see any point in continuing upwards with 2TB drives.

My issue is I've seen people saying that with some drives, given their error rate and size that RAID 5 isn't good enough? It's all leaving me a bit confused as I understand the basics of RAID just not the reasons why I would need to move up to RAID 6 or similar when increasing the drive size.

What would people suggest as a way forward? I was thinking 4TB drives this time. Essentially this would mean I could create a 3 drive RAID 5 array of the same size as the existing array and then retire the current RAID 5 array leaving me 5 free ports to expand into.

I know I can do RAID migration on this card but I'm just looking at the best way forward for me.

It is used for video/audio/pictures storage for streaming around the house, as well as allowing me to run several websites locally for development.

Any help/suggestions greatly appreciated

Cheers,
Paul.

[kalniel]

As far as I can tell, it's just maths - larger disk = higher potential for errors in one disk assuming a constant error rate per number of sectors. Balancing that though is the reduced chance of mechanical failure by having fewer disks. Typically you are moving up disk size because your IO is increasing, so you are read/writing more sectors as well, also increasing error rate. If your error rate increases then you might need to consider moving to a more robust RAID mode.

But I don't think you need to worry - it doesn't like like your IO is increasing, you're just consolidating disks. Take account of the longer rebuild time for larger volumes and if that's no problem I don't see why RAID 5 wouldn't still be OK for you - I'm not an expert however.

[DanceswithUnix]

It comes down to this. If a drive fails then you need to rebuild the array with a new drive. During that time the array is vulnerable to losing data if another drive fails during the rebuild. The bigger the drives, the longer the rebuild time, the greater the risk.

Modern systems like raid-z only rebuild the live data rather than every single block, so faster so a bit less risk.

Raid isn't a substitute for backups, so you have backups right? If so, you can probably wear the risk of a single parity disk.

[probedb]

Thanks for the replies folks. I may go with RAID 5 with larger discs next, then I can ditch the smaller drives and possibly migrate to RAID 6.

No, these are not backups!

[DanceswithUnix]

No, these are not backups!

But the data is backed up somewhere, or you don't care if you lose it?

RAID ups the odds of data being kept intact on a computer due to mechanical drive failure, but it offers nothing against things like accidental deletion or the OS suddenly losing its marbles & deciding it likes to write random data to random places really fast.

I'm told if someone uses a pneumatic drill next to a running industrial grade storage rack, even raid 6 doesn't help

[shaithis]

There was a guy who posted something years ago about RAID reaching a point of being useless due to failures and rebuild times (in fact I found the article)....I think he didn't take disk and controller speeds into account It did go viral at the time and I still hear people mentioning it occasionally.

I wouldn't worry about it though, I look after a lot of storage including a few 24 x 4TB disk RAID6 arrays. Rebuilds and expansions take a while but shouldn't be an issue unless you use Seagate drives!

[DanceswithUnix]

....I think he didn't take disk and controller speeds into account It did go viral at the time and I still hear people mentioning it occasionally.

Better disk and controller speed just get you to the point of failure faster

The thing is here, if the capacity of a drive has doubled, then the unrecoverable error rate (ie number of bits read per error) must halve to compensate.

Edit to add: This is where spending the big bucks helps. If you use consumer WD RED drives (like I do), you get a rating of 1 error in 10^14, if you use the enterprise grade RE drives, you get 1 in 10^15.

http://www.wdc.com/wdproducts/librar...879-771444.pdf
http://www.wdc.com/wdproducts/librar...879-771442.pdf

So that is a magnitude better odds of each drive giving a successful rebuild.

[probedb]

But the data is backed up somewhere, or you don't care if you lose it?

RAID ups the odds of data being kept intact on a computer due to mechanical drive failure, but it offers nothing against things like accidental deletion or the OS suddenly losing its marbles & deciding it likes to write random data to random places really fast.

I'm told if someone uses a pneumatic drill next to a running industrial grade storage rack, even raid 6 doesn't help

Well backups are a separate conversation hence why I just talked about RAID

I backup the bits I need to be backed up

[jimbouk]

I'm told if someone uses a pneumatic drill next to a running industrial grade storage rack, even raid 6 doesn't help

I kinda want that quote on a mug or t-shirt now