Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[translux]

I'm sure many of have seen this.
Deciding on the right drives especially for an array is not easy.
The findings here are representative up what we have seen. Of course on a much smaller scale.
As reflected I'd stay away from Seagate. They have some serious bad mojo. Hero to zero.
So far HGST drives for us have been bullet proof.

https://www.backblaze.com/blog/best-hard-drive/

https://www.backblaze.com/blog/wp-content/uploads/2015/01/blog-drive-failure-by-manufacturer1.jpg

[mwillems]

There are some serious problems with that chart.  If you look at the data presentation above the chart, you can see that their population of seagate drives is drastically older on average than the other disk populations, and there's no attempt to normalize the data for hard-drive age.  The seagates are probably still worse, but there's no way to tell from that data.

The WD Drives in the chart are on-average less than a year old.  They won't be comparable to the seagates, which are on average 3-5 years old (i.e. when hard drives start failing).

This has been a persistent problem with their data presentation in their past posts, and I have no idea why they don't try and present more meaningful data as it could be a gold mine for researchers.  In last year's they at least offered a 36-month survival rate, which is more meaningful.

Goofy methodology is a problem that seems to afflict almost every "study" of harddrive failure I've seen, with the exception of google's amazing white paper: http://static.googleusercontent.com/media/research.google.com/en/us/archive/disk_failures.pdf

But that's 8-years old now  

[translux]

Hum maybe I'm interpreting the data differently.

The 3TB Seagates with an incredible 43% failure rate are younger than the HGST.
Years ago Seagate was the only choice. Today HGST is the best IMHO
I have bins of dead Seagate drives if anyone needs any.

[Hendrik]

An important factor to consider here is that Backblaze uses Desktop drives in a extremely high density server configuration.
At their configuration, they suffer terribly from vibrations, which is something a desktop or even home server setup will generally never have to worry about.

Keeping that in mind, any data obtained by them should be taken with a pound of salt.

[translux]

Given the HGST drives are in the same environment and are also desktop drives the comparison to the Seagate's is relevant.

As I mentioned the data parallels my personal experiences.
I will say that the Seagate Constellation series have fared much better.
Given their cost I don't think they are practical for consumer use.
Also I would not hesitate to recommend an HGST desktop drive at 1/2 the cost of a Seagte enterprise.

The common misconception of "I'm using a raid so no big deal if a drive fails" is in fact a huge deal. Even with high end hardware and software rebuilding a large array takes a significant time investment and is a crap shoot at best. With a successful recovery things are really never quite the same. I've moved to mirroring and when there is a failure promote the mirror to primary and start from scratch with new drives for the mirror.

I'm bitter from the pain I have experienced due to their drives failing.
How the tables have turned. Deathstar as they were commonly know now offer superior reliability with Seagate once THE industry standard at the bottom of list.

[mwillems]

Hum maybe I'm interpreting the data differently.

The 3TB Seagates with an incredible 43% failure rate are younger than the HGST.
Years ago Seagate was the only choice. Today HGST is the best IMHO
I have bins of dead Seagate drives if anyone needs any.

I think it's safe to draw the conclusion that that particular model of seagate drive (7200.14) is a problem, and that Deskstars are more reliable on average than the other drives in the chart.  Other than that there's very little meaningful information in the table.  

The WD info is meaningless for comparison because all the drives are too young, and the two seagate lines above the two you quoted are also meaningless because they're the only drives on the chart that are almost five years old which is where things get dicey for drives.  

To be clear, I'm not a fan of seagate at all (seagate does actually have serious reliability problems); my point is that the chart prepared from that data table is misleading and kind of irresponsible because it compares 5 year old drives to 1 year old drives without any indication that's what's happening.  

That doesn't mean seagates aren't unreliable or that the 7200.14 isn't a total lemon; it just means they didn't do a very good job of supporting that conclusion.

They could've just adjusted the data for drive age, or split them into buckets, or weighted it, or done any number of things to present meaningful information.  It looks like they recently published their actual raw data set, so hopefully someone will actually do some rigorous analysis at some point.

[Hendrik]

Given the HGST drives are in the same environment and are also desktop drives the comparison to the Seagate's is relevant.

If you want to run it in the exact same conditions, sure. :p

I'm not a Seagate fan by far, but I strongly dislike all the weight and importance people put on those Backblaze stats. Their usage is very special and very far from typical consumer use, and they are using drives which are not designed for this.
Of course its nice when some other drive type doesn't fail in the same circumstances, however its hard to conclude that it would fail in equal numbers in "normal" usage.

[translux]

While we're on the subject do you think the drives designated for specific use, NAS, Security etc. is legit or a marketing exercise?

[Hendrik]

I buy NAS drives because they come with a longer warranty - and the manufacturer claiming they are designed to run 24/7 is at least better than claiming its supposed to run for 8 hours a day max (business hours).
"Security" drives are designed for 24/7 write operations, like for operating a security camera which is constantly writing to it. I'm sure they can make some optimizations for these use-cases, but how big of a difference that makes is for them to know and us to guess.

The NAS drives are usually not much more expensive, so the extended warranty alone is worth it for me.