Drives for Home-built NAS

[rfielder]

Just to be clear - I am comparing spec sheets here ONLY. No flaming or religion wars in the Seagate vs WD battle, please. Had I the cash, I would be going enterprise, maybe choosing neither brand.

Looking at 8Tb drives for the NAS I am configuring (in my mind and on paper).

Looking at the specs for the WD Red and the Seagate Ironwolf, I see that Seagate seems to have pulled ahead.

All Seagate Ironwolf drives from 6Tb and up are 7200rpm. All WD Red at all sizes are 5400rpm. That is interesting - I have some older Ironwolf, and I would have sworn that they were 5900rpm.

Something I noticed is that WD Red drives are listed as <1 in 10^14 for "non-recoverable errors per bits read".

Seagate lists the rate as 1 in 10^15 for the 8Tb Ironwolf. I did not check other sizes.

That error rate is kind of a huge difference, isn't it?

My personal experience - I have a WD Gold 2Tb drive in my server that died in just over 1 year. It has a 5 year warranty. I have lost both Seagate and WD over the years, and have some really old drives that seem happy to sit there and wait to galactic entropy to end their run.

 

[Mlovelace]

That error rate is kind of a huge difference, isn't it?

It is in fact an order of magnitude different. ;) Though they are merely probability estimates and tend to be quite conservative when borne out over time.

Both drives support TLER and the performance difference, as negligible as it is, won't make a difference in home use. So, buy which ever drive is the better price.

 

[rvassar]

The 7200 RPM drives will likely not give you much realizable performance gain, unless you have multiple 1GbE networks, or 10GbE networking in your house. They draw more power, and produce more heat. The heat leads to premature failure.

 

[Heracles]

I am using IronWolf here because they were better priced when I bought them.

I agree with MLovelace in that for a home usage, you will not really see the difference between the two. The pool layout, the amount of RAM, the CPU, the network and more are million time more important.

 

[rfielder]

The 7200 RPM drives will likely not give you much realizable performance gain, unless you have multiple 1GbE networks, or 10GbE networking in your house. They draw more power, and produce more heat. The heat leads to premature failure.

I am sorry, but my experience has been just the opposite.

When new generations of drives are released, they tend to be more efficient.

The Seagates drive the same power as the WD, according to the spec sheets, so they are getting faster without pulling more power, which may imply no additional heat.

 

[rfielder]

I am using IronWolf here because they were better priced when I bought them.

I agree with MLovelace in that for a home usage, you will not really see the difference between the two. The pool layout, the amount of RAM, the CPU, the network and more are million time more important.

I agree - but for the same price, might as well go for the faster drives.

The bit failure rate is what really caught my eye. If this has any meaning in the real world we live in, it could mean a LOT less opportunity for two drives to fail at once. Like, about 10 times less chance!

I would not want to pin my reputation on it, but unless Seagate has abandoned all credibility, it may have meaning.

I noticed that most of the enterprise disks also have the 10^15 for (I think) both brands. Moving that reliability from enterprise to NAS is a good thing.

 

[rvassar]

I am sorry, but my experience has been just the opposite.

Yes it's possible newer 7200 rpm drives may be more efficient. Hasn't panned out for me so far. I have a mixed pool, and the 1 year old 7200 rpm drives run 4 - 5 deg/C hotter, sitting in front of the same 120mm fan. I was referring to what's needed to saturate 1GbE. From a remote client, a saturated pipe is a saturated pipe.

The bit failure rate is what really caught my eye. If this has any meaning in the real world we live in, it could mean a LOT less opportunity for two drives to fail at once. Like, about 10 times less chance!

Now here I'm going to suggest you go read what that bit error failure rate actually is. It's something of a stretch to go from 1^14 to 1^15 to... "less opportunity for two drives to fail at once". That statistic is unrecoverable read error, not drive failure. In ZFS an unrecoverable read error gets corrected by ZFS checksum and the block remapped. This is why we run scrubs. That doesn't mean the drive is "bad", or even that it will get marked as failing. Now the 10^15 number does influence the odds of getting two events in the same parity stripe, so yes, it is better. But it's easy to overstate the significance.

I have a drive that has three bad sectors. They were marked bad around the 7000 hour mark, a few hundred hours apart. The drive has 38k+ hours on it now, and hasn't had any more errors.

 

[rfielder]

Yes it's possible newer 7200 rpm drives may be more efficient. Hasn't panned out for me so far. I have a mixed pool, and the 1 year old 7200 rpm drives run 4 - 5 deg/C hotter, sitting in front of the same 120mm fan.

Thanks - that is good to know!