Large volume HDDs bad idea?

A few years ago someone told me that it’s better to have 2 x 6tb drives instead of 1 x 12tb because larger volume HDDs tend to fail faster than smaller ones. Is this still true? I don’t have much space in my current enclosure and I’d prefer to not have to buy a new case too. Thanks.

1 Like

Not really as now you’re rolling 2 dice rather than 1 now

1 Like

I’m rolling in 6 x 16tb 2x14tb 3x12tb and 5x10tb none have failed thus far.

1 Like

So long as you’re not trusting important data to a single drive I don’t think capacity really matters that much.

Just get an external drive if you need to and do backups to that.

4 Likes

I don’t think that has ever been true. Doubling the number of drives, you double the odds of experiencing hardware failures. The manufacturer’s spec sheet on a given hard drive should show a MTBF figure for the specific drive, and you certainly won’t see MTBF figures going down to HALF as capacities increase. In fact, you’re likely to see MTBF figures going UP with newer drives, instead.

This advice was likely a misunderstanding of either performance issues, or redundancy (RAID).

As @GigaBusterEXE says two drives is twice the chance of failure.
However it’s more complex than that. You’d need to compare the reliability of the two actual drive choices. With with twice as much storage the larger drive will naturally be more likely to break unless they mitigated for that.

Ideally buy two or more drives and run them in RAID or RAIDz (ZFS). You should expect a HDD to break.

3 Likes

I’ve been running 10 x 12TB WD Red’s in my RackStation for well over 3 years without a single issue. None have failed, and they have been on non-stop other than required reboots.

I pay £8.70 each for my 3TB SAS drives off ebay. I get some failures in the Seagate, thankfully the Hitachi have been fine.

You could look at it another way.

*Numbers made up for effect
If (Big if) 5% of drives die each year, over 20 years, pretty much all drives will be dead, but the average age is 10 years.
Buy 2 drives, then there is 10% chance one of the drives will die each year. (or 25%. whatever. the number is higher than single drive)

But When a drive dies, you only loose half the data, and still have the other half.
And not all failures mean absolute loss.

RAID is not a backup, but it can buy you more time to replace a dead drive.

And, All Drives Die.

(I personally prefer more, smaller drives, and RAID them)

2 Likes

6vgs4l

6 Likes

I lub it. Muchly!

1 Like

Smaller drives is relative. I took out 8x 143GB 2.5" SAS drives and replaced them with 8x 600GB drives. Am I still using small drives?

I have a TrueNAS with 8x 3TB drives I bought off ebay for £87 for a box of ten. I regard them as large but these days they are small.

I think the point is it’s easy to build a system with two 20TB drives but hard to build one with eight 3TB drives. but the result is about the same.

The dual drive system has greater redundancy for similar capacity.

Sorry to necrobump this, but I was thinking about this and something doesn’t make sense to me. If say I have all of my data split between two hard drives and one of the HDDs die I only lose half of my data, but if I have all of my data on one hard drive and it dies, I lose all of my data. In that sense how is having more than one drive riskier than one drive? Thanks.

Backups. Backups. Backups. It doesn’t matter if you have a calendar events to make regular backups of your drives.

For years I just used large drives. When my 3TB was close to full, I upgraded to ta 4TB and used the 3TB as a backup. When the 4 went over the capacity of the backup, I went up to a 6TB.

I current have an 18TB, 14TB and 12TB for primary storage (separated by data type) and a 14TB, 12TB and 8TB as backups for all of those.

Your odds of each drive failing stay the same (let’s say: 1% chance each year), but now you have two of them so your odds of losing (half) of your data has doubled (so: 2% chance each year).

Besides, the talk is usually about multiple drives in the context of RAID… Specifically RAID-0 of two drives versus a non-RAID single drive. In which case, either drive failing causes loss of everything. Not restructuring how you store data.

Yeah but the likelihood of both hard drives failing at the same time is extremely low. So chances that half of my data will survive on two hard drives are higher if data is split between two hard drives, no? Also I am not sure that probability compounds like that - this would imply that if I have 100 hard drives, there’s 100% I will lose 1/100th of my data, 99% I will lose 1/99th of my data etc. basically guaranteeing that I will be losing more data if I have more disks.

RAID0 is out of the question, so that’s not a real concern for me.

@rcxb I just edited this in, and will edit it out again in a bit,

House always wins. Odds stacked against you, but isn’t it more like 18/38? 0&00 green as well as 18 of the other colour?
Hard drive percentages are not hard numbers, or the producers could well bevkying of course… But roulette is standardised? Maybe

I think you have the math mostly right.

But you are only likely to loose a portion per year,as the drives die.
It is not exactly linear, often a bell curve.

But having more drives, spreads the loss out more.

Backblaze does reports on it’s drives, so you can see a 2-8% annual failure rate is about right.

So 10 people with 1 drive each, can be all fine for 3 years, then one person looses all their data, then nothing for 3 years then 4 people loose their data then nothing, then nothing , then one more .

I mean to say, snall sample size, has 100% working or not working. Averages are for large numbers.

A 1% rate, means your disc could last 100 years. Or a quarter of a year, and someone else might last 100 years

I might be wrong, but I think I remember it goes like this:
100 drives, simplyfied anual failure rate 0.01 (1%).

Probability to lose one drive in first year: 100 * 0.01 = 1 → 100% you lose one drive in first year.
Probability to lose 100 drives in first year: 0.01^100 = 1E-200

However the failure rate is not constant over time. I think it’s higher at the start (potential manufacturing defects) and towards the end (age), and lower in the middle.

3:2:1 Rule

Have 3 copies of your data, stored on 2 different mediums, with one copy secure off site.

So a 20tb ZFS array, with a backup on a 20tb WD drive, with a copy on a 20tb SEA drive stored at a friends or family members house meets this requirement.

The chance of you losing all your data is exactly the same with one drive or two. 1%/1=1%. 2%/2=1%. The fact that the two drives likely won’t fail at exactly the same time is the reason behind RAID, but it doesn’t help you at all in your scenario unless there’s something you’ve failed to mention… What will you do after losing half your data, to prevent losing the drive with the other half later?

I’m afraid so.

Brain teaser for you: After the roulette ball lands on red 4 times in a row, what are the odds at the next spin it will land on black?

You should really think it through and try to guess.

Answer: Still 50/50, just like always.