Reliable External Hard Disk questions and recommendations

There's a huge difference in a Google server farm and the OP's presumed usage where I am sure that a drive failure doesn't have the catastrophic effect on the whole operation that a drive failure would have on an individual.
Are you suggesting that the likelihood of failure is proportional to the size of the catastrophe it would cause?
The point was about the usage. Google stresses drives like no individual can.
So.... the fact that Google's supposedly highly stressed consumer drives didn't fail at any significantly greater rate than it's enterprise drives means... what?

You keep telling us that you've never had problems with enterprise drives. Bully for you! But Google (and other companies that have reported similar results) have gone through a LOT more drives than you have, and frankly we'd be crazy to take your word for it over theirs. You can see that, right?
 
Last edited:
It is quite clear that you either have not read the paper yourself (rather depending on some other internet troll to do your thinking for you) or you do not understand what you've read.
No, I think that kelpdiver and I (presumably the "other dude") understand the literature a lot more thoroughly than you do.

You didn't understand how Backblaze designs its approach to modular, replaceable storage units and thought they bought enterprise disks for magical reliability.

You referenced a lite-reading ZDNet summary instead of the actual Google/UoT report, which causes you to then go and and...

... Confuse the 2016 report on SSDs with the 2007 report on hard drives.
The point was about the usage. Google stresses drives like no individual can. Yet, we (consumers) all have had consumer drives fail. I, as an IT consultant, has never had an enterprise drive fail in my own computer or a customer's. Of course, I am going to recommend what I know.
And it's become apparent that this is willfully not much as all, so of course you're going to continue to promote an unnecessarily expensive, yet similarly failure-prone solution.
I suppose you read what you want, but I said early on cloud storage is the most reliable and safe - and let me add this - for the individual. But again, the one ugly factor that will raise its head on this idea too is the cost of it.
And how is this is relevant to the selection hard drives? We all know what cloud storage/backup is, and the majority of regular posters in PC Talk seem to be using it already. But nobody brought it up, because it has nothing to do with hard drive reliability.

Hm... why did even I bother writing this. From experience, the target is neither going to read or be able to understand it.
 
Last edited:
It is quite clear that you either have not read the paper yourself (rather depending on some other internet troll to do your thinking for you) or you do not understand what you've read.
No, I think that kelpdiver and I (presumably the "other dude") understand the literature a lot more thoroughly than you do.
It does not appear that you do.
You didn't understand how Backblaze designs its approach to modular, replaceable storage units and thought they bought enterprise disks for magical reliability.
I understand the Backblaze approach quite thoroughly. This isn't the first thread where an inexperienced person throws Backblaze at me as if to prove their choice of $99 HDs from Best Buy are ok for archival.
You referenced a lite-reading ZDNet summary instead of the actual Google/UoT report, which causes you to then go and and...
So?
... Confuse the 2016 report on SSDs with the 2007 report on hard drives.
I am surprised you can spell computer. I know this push-back is all about defending your cheap built bedro9om computer but c'mon grow up.

The point was about the usage. Google stresses drives like no individual can. Yet, we (consumers) all have had consumer drives fail. I, as an IT consultant, has never had an enterprise drive fail in my own computer or a customer's. Of course, I am going to recommend what I know.
And it's become apparent that this is willfully not much as all, so of course you're going to continue to promote an unnecessarily expensive, yet similarly failure-prone solution.
I am going to recommend what I know works. If you are an adult (and I have my doubts at this point), you'll recommend what you recommend and the OP will make his decision.
I suppose you read what you want, but I said early on cloud storage is the most reliable and safe - and let me add this - for the individual. But again, the one ugly factor that will raise its head on this idea too is the cost of it.
<snip further foolishness>
 
OK - to answer most of your questions, mine are the standard 3.5" variety. My current WD external drives are 2 Terabytes in size. I'm looking to move up to anywhere from 4-8 TB in size.

It's interesting to find out that the only OEM drive makers remaining at Seagate, WD, and Toshiba. Makes the narrowing down very easy.
And probably, your 4-8Tb requirement will narrow the possibilities even further. :-D
Reliable will narrow the possibilities further.
I presume that you saw this WD announcement
If the OP wants 'reliable', he'll pursue this model. It comes in an 8 TB version. Enterprise drives don't fail often if at all.
I was curious what customer reviews on the 8TB gold "Enterprise" drive were like so I checked on Amazon.

15 units sold

1 guy had two 8TB golds fail

another guy also reported a defective drive

another guy reported loud clunking noises and returned the drive for a Seagate

This is based on a sample size of 15, not too much to go on and up there with anecdotal stories. But as such, it certainly puts into question this drive's superiority to other drives folks here use and any claims of invulnerability (don't fail often if at all).

Also reported that units are made in China

Also reported that some folks are not even getting a 5 yr warranty

Instead of sweating failure rates, much more productive, and much safer, to have a really good backup plan.
 
OK - to answer most of your questions, mine are the standard 3.5" variety. My current WD external drives are 2 Terabytes in size. I'm looking to move up to anywhere from 4-8 TB in size.

It's interesting to find out that the only OEM drive makers remaining at Seagate, WD, and Toshiba. Makes the narrowing down very easy.
And probably, your 4-8Tb requirement will narrow the possibilities even further. :-D
Reliable will narrow the possibilities further.
I presume that you saw this WD announcement
If the OP wants 'reliable', he'll pursue this model. It comes in an 8 TB version. Enterprise drives don't fail often if at all.
I was curious what customer reviews on the 8TB gold "Enterprise" drive were like so I checked on Amazon.

15 units sold

1 guy had two 8TB golds fail

another guy also reported a defective drive

another guy reported loud clunking noises and returned the drive for a Seagate

This is based on a sample size of 15, not too much to go on and up there with anecdotal stories. But as such, it certainly puts into question this drive's superiority to other drives folks here use and any claims of invulnerability (don't fail often if at all).

Also reported that units are made in China

Also reported that some folks are not even getting a 5 yr warranty

Instead of sweating failure rates, much more productive, and much safer, to have a really good backup plan.
 
Instead of sweating failure rates, much more productive, and much safer, to have a really good backup plan.
Which is what I suggested 4 days ago.


You'll probably disagree. I am surprised someone didn't pick up on that and make note of the fact that my primary backup plan is most likely going onto consumer drives.
 
For years, it has been my experience that Western Digital external hard drives have shown fairly long term reliability. In my case (and I realize everyone has had different experiences), I've had less luck with Seagate models (all of them having crashed and I've had to discard them).
1. WD and Seagate both make millions of HDDs a year. They sell to consumers and they sell to enterprise storage farm operators, and everything in between.

2. Anecdotal failure-rate experience is meaningless data.

3. Both vendors make a wide range of drives of different capacities, quality levels, performance, etc etc.... Don't compare apples and oranges.

4. HDDs will fail at a rate of 1% a year, more or less. Cherry-picking will not reduce that failure-rate significantly. So your backup scheme should include redundancy on several drives, in more than one location.
 
I am surprised you can spell computer. I know this push-back is all about defending your cheap built bedro9om computer but c'mon grow up.
Growing up would entail fewer straw men on your part. My last "bedro9om" computer order was over $20M. Many of the machines that I've had to replace drives on in the past cost well over 100k each, with unit prices on the drives at over a grand each.

And you continue with this false dichotomy. No one has asserted: 'you should only buy the cheapest drive available.' No, the very clear statement, in contrast to your beliefs based on extremely thin experience, is that spending more for 'enterprise' class drives will not lower your risk of a drive failure.

You're a minnow swimming with sharks.
 
I was curious what customer reviews on the 8TB gold "Enterprise" drive were like so I checked on Amazon.

15 units sold

1 guy had two 8TB golds fail

another guy also reported a defective drive

another guy reported loud clunking noises and returned the drive for a Seagate
Every new generation of capacity bursting units seem to have a rough start. Getting that new level generally requires a first gen model with additional platters (more failure points) or higher density (more potential for trouble). After it matures, then the tech filters down to the 'lower' capacity drives, so the 4tb drive that once had 5 platters might be down to 3.

This likely pushes us to the 4s or the 6s as the right choice. Staying far away from the 10s and 12s. The cost premium for the 8s has large dissapated, but the history still seems unstable.
 
A digression, but I'm curious as I haven't been in a seriously large scale enterprise environment for fifteen years now. Back then the emphasis was all SAN/NAS with fibrechannel connects, and site redundancy was heavily reliant on centralised storage racks being replicated to remote sites via dark fibre. This was high performance, highly reliable, but staggeringly expensive to set up and maintain.

What's the storage approach used in your environment? As it sounds seriously large. Centralised, distributed (I'm very curious as to how Hadoop or similar approaches are common these days), or the primitive but simple disks-in-individual-servers. I'd be guessing not the last one if only because paying people to run up and down aisles doing hard drive replacements can't be a good way to spend money.
Like many companies with rapid growth and high acquisition rates, we have inherited a hodgepodge of storage products. That leads to the dilemma of maintaining many, or going through the work to try to consolidate.

DB still lives on SAN, and that has been the primary data store historically. But need to provide customers broader storage on our premise lead to a product based on 2U skus with 25 drive bays, as well as an hbase solution, and ceph is the faddish thing now.

The trend line does seem to be heading towards massive variation of RAID - lots of disks on lots of increasingly vanilla systems with 3 copies of everything. More the swarm mentality than the big iron of the 2000s. Less prone to complete failure. Unfortunately for me, more difficult to maintain (patching, in particular).
 
4. HDDs will fail at a rate of 1% a year, more or less. Cherry-picking will not reduce that failure-rate significantly.
This, of course, is what BackBlaze figured out, and hence they buy primarily on price.

It seems to be working for them. They replace drives every day but their RAID and rack design makes it no biggie.

For consumers, it's a little different with just one or a few drives in play. Failures will be rare but one MUST plan for them with a suitable backup protocol. If you buy cheap, the odds still strongly favor a cost effective outcome.

This is bad news for just one audience however: disk drive manufacturers and their razor thin margins.

BTW, although many/most drives are manufactured in China/Asia, some of the most critical components are not. WD, for example, make their heads about a mile down the road from my home in Fremont, CA.
 
Like many companies with rapid growth and high acquisition rates, we have inherited a hodgepodge of storage products. That leads to the dilemma of maintaining many, or going through the work to try to consolidate.

DB still lives on SAN, and that has been the primary data store historically. But need to provide customers broader storage on our premise lead to a product based on 2U skus with 25 drive bays, as well as an hbase solution, and ceph is the faddish thing now.

The trend line does seem to be heading towards massive variation of RAID - lots of disks on lots of increasingly vanilla systems with 3 copies of everything. More the swarm mentality than the big iron of the 2000s. Less prone to complete failure. Unfortunately for me, more difficult to maintain (patching, in particular).
I hadn't even heard of Ceph. More downtime reading for me then.

My environment has the main databases in RAM (obviously quite specialised databases of limited size) with only relatively static data stored on disk. Nothing like that number of physical servers either, and an organically grown infrastructure. So whilst patching is a pain, it's still relative homogeneous.
 
Last edited:
Instead of sweating failure rates, much more productive, and much safer, to have a really good backup plan.
Which is what I suggested 4 days ago.

https://www.dpreview.com/forums/post/60151357

You'll probably disagree. I am surprised someone didn't pick up on that and make note of the fact that my primary backup plan is most likely going onto consumer drives.
Yes, you keep bringing up that point like it is an overlooked piece of genius. Backup threads pop up at least every week. Backups are already universally recommended in PC Talk.

The main point of contention now is that enterprise drives are a foolish proposition for the home user, or just the usual cost of doing business for the enterprise (we use enterprise drives, they're bundled into the annual maintenance costs of the server manufacturer).

Recommend what you want. But if you recommend arrant nonsense, expect to be called out on it by people who know a lot more than you. 20 years in IT doesn't mean squat if you haven't kept learning throughout those twenty years.
 
Last edited:
This is bad news for just one audience however: disk drive manufacturers and their razor thin margins.

BTW, although many/most drives are manufactured in China/Asia, some of the most critical components are not. WD, for example, make their heads about a mile down the road from my home in Fremont, CA.
That's an interesting piece of trivia, which in hindsight it does makes perfect sense. Ultra precise gear probably isn't made anywhere in Asia save for Japan, and Japan is expensive and a bit weird to do business in.

A comparable, but far more mundane, example would be the ball tips that go into pens. These things are mostly made in Switzerland.

http://www.businessinsider.com/china-has-finally-figured-out-how-to-make-ballpoint-pens-2017-1
 
Last edited:
My environment has the main databases in RAM (obviously quite specialised databases of limited size) with only relatively static data stored on disk. Nothing like that number of physical servers either, and an organically grown infrastructure. So whilst patching is a pain, it's still relative homogeneous.
ksplice? I'm not personally an advocate - been trying to break role owners of the mentality that long uptimes and avoiding reboots is a good strategy over a more resilient service. But Oracle RAC is a delicate beast and the db guys are going to try that route out.

We suffer mostly from massive databases due to the original route of storing everything there. But soon moving to a model of smaller chunks.
 
My environment has the main databases in RAM (obviously quite specialised databases of limited size) with only relatively static data stored on disk. Nothing like that number of physical servers either, and an organically grown infrastructure. So whilst patching is a pain, it's still relative homogeneous.
ksplice? I'm not personally an advocate - been trying to break role owners of the mentality that long uptimes and avoiding reboots is a good strategy over a more resilient service. But Oracle RAC is a delicate beast and the db guys are going to try that route out.

We suffer mostly from massive databases due to the original route of storing everything there. But soon moving to a model of smaller chunks.
We're certainly not advocates of long uptimes! RAM gets written to disk on a scheduled basis so reboots are not an issue.
 
Last edited:
We're certainly not advocates of long uptimes! RAM gets written to disk on a scheduled basis so reboots are not an issue.
I miss the good old days of core memory when you could power a computer down, come back the next day, power it back up again, and it would continue on right where it left off...
 
Last edited:
We're certainly not advocates of long uptimes! RAM gets written to disk on a scheduled basis so reboots are not an issue.
I miss the good old days of core memory when you could power a computer down, come back the next day, power it back up again, and it would continue on right where it left off...
... after entering the boot-up instructions on the front panel switches... I still remember the PDP-8 that I learned Assembly programming on, with its paper-punch tape and teletype terminal.
 
I saw a functioning PDP-1 (presumably the only one in the world) a couple of months ago and was amused by the ticker tape program loads. People were playing a precursor to Asteroids on it (somehow they simulated gravity too). Here's a photo of the screen.

7f140929b9e94be7ab53a57f656cceca.jpg
 
Last edited:
I saw a functioning PDP-1 (presumably the only one in the world)
There are still many in existence. DEC sold them for 10 years.
a couple of months ago and was amused by the ticker tape program loads. People were playing a precursor to Asteroids on it (somehow they simulated gravity too).
Real-time interaction with the user during the execution of the program was the revolutionary contribution of the PDP-1 designers and programmers. Until then, computers ran programs that would load, execute, spit out results and stop.
 

Keyboard shortcuts

Back
Top