Estimating Life Expectancy of Hard Drives is Tricky Business

33

Comments

+ Add a Comment
avatar

Mark17

"Overall, I argue that the enterprise drives we have are treated as well as the consumer drives. And the enterprise drives are failing more."

I don't understand how the author comes to this conclusion after stating that the drives were in completely different environments.

I find it hard to believe that consumer drives are more reliable than enterprise drives. In order to test whether enterprise drives are more reliable than consumer drives, you would need to test them under the exact same conditions and have the exact same number of reads and writes applied to both. Not only that, but if you're going to generalize "enterprise" and "consumer" drives, you would also need to test many different makes and models of both types of drives. It could be possible that one make and model of a consumer drive would outperform a specific make and model of an enterprise drive, and vice versa. To come to any conclusion, a controlled test would need to be performed.

The Backblaze statistics tell us hardly anything as you can't directly compare the results. What they've done is no different than comparing peanut to a walnut. With the walnut getting run through clothes dryer until it cracks and the peanut getting hung out on a clothes line until it cracks. How do you know which one has the more reliable shell based on those results? I would tell you, but the squirrels stole all my peanuts.

avatar

Belboz99

When shopping for the best hard drives on New Egg, I used the "1 Egg" rating as a means to gauge infant mortality in drives.

Now, I highly doubt the amount of 1 Eggs is directly propotional to the infant mortality rate, mostly because people usually are more inclined to give bad feedback on a failure than good feedback on a success, but...

What I did notice was a steep increase in 1-Egg reviews once the 1TB threshold was crossed. By-and-large, 500GB and 750GB HDD's had around 1/2 the percentage of 1-Egg reviews as the 1TB and 2TB models. Sub-1TB are typically 5-10% of reviews, beyond 1TB it's typically 10-20%.

Again, that's not likely the actual rate of infant-mortality, but you can easily see which drives are likely to have the highest, and on-average, that's the 1TB+ range.

Dan O.

avatar

Gikero

I've noticed more failures in 1-3 year old PCs, that have 500GB-2TB drives. It has gotten to the point that I am personally worried about my data on anything bigger than a 320GB. I am sure not all large drives are bad, and I also think that is partly because most of said PCs were budget systems. They use the cheapest components.

avatar

tnguy

When Backblaze cites "years", do they mean literally that of the drives they use 24/7, 26% fail at around 4 years? Would that mean if a user typically runs their HDD 12/7 the outcome would be different?

avatar

Richardbs

Is there a failure rate on SSD's? I've been curious on how reliable they are

avatar

poee

As these comments demonstrate, anecdotal evidence is irrelevant.

avatar

Hey.That_Dude

Query: What separates anecdotal from hard/concrete?

Is it the "volume" of data collected? If so I could compile millions of users data into a sufficiently large study to draw hard facts from.

Or is it that people let their emotions get in the way and don't provide hard timing and confirmed defects? Understandable. However, you'll still be insulting all of the people who actually provide decent data.

avatar

Dark_wizzie

I don't understand what you're saying. Of course it's the size of the data collected. You could in theory go around and collect 100,000 data points on hard drive failure chance but you're not going to. Any individual stories of "This brand sucks" or "don't trust this" doesn't mean anything.

avatar

Hey.That_Dude

True; but on this page, as of this posting, there is EXACTLY ONE of those ' "This brand sucks" or "don't trust this" ' reviews. Everyone else has just given a brief run down of how their hard drives, in general, have fared. So, I was giving a little poke to the people who like to poke everyone for posting general conclusions drawn from 1ST HAND DATA. (Some of the best data there is)
And while no ONE story of "OMG! this brand sucks." should be taken too heavily. A THOUSAND accounts of "OMG! This brand bites." might be an indication of quality control issues or just bad tech.
THAT, is the point I'm making. The difference between anecdotal and hard facts is the VOLUME. So, stop discouraging people from posting facts (not opinions, the fact's the important part) and instead ENCOURAGE THEM so that we have more data to more accurately see real world patterns.

avatar

BlazePC

Amen!

avatar

John Pombrio

Cripes, everyone here are children (of course I always knew that). 61. I started working for HP as field test and measurement system support. did that for 25 years.
Hard drives back in '81 were 10MB, two platters, each the size of a medium pizza, they were in a box the size of a dishwasher. The heads were exposed, just using highly filtered air to keep dust from crashing them. The first sealed drive that HP made was 20MB and seventy pounds.

I have seen hundreds of broken hard drives. I never found much correlation between heat, dust, or environment causing premature failures. Only dead fans or blocked fan enclosures with those damn dust filters that were never cleaned would kill stuff inside. At home, my PC computer builds starting in '92 or so has seen a remarkable improvement in hard drives. Today, no real issues with any brand, size hard drive,or spindle speed. I used to buy Hitachi drives, now Seagate Barracudas.

Lately, most disk drives failures I have seen at home since I retired 10 years ago are progressive: bad sectors or intermittent op sys failures. Three to four years is about right. ANY drive that I have ANY suspicions at are immediately taken out and recycled. Overall, the drives these days are remarkably good with no DOAs that I can remember in the past 15 years.

BTW, I have had one SSD failure out of the eight I own.

avatar

Gikero

I'm too young to know myself, but I love computer history. Dishwasher sized 10MB drives. I wonder how much power it sucked down.

avatar

The Mac

quite a bit more than a dishwasher

lol

avatar

BlazePC

You're an anecdotal anomaly John. Consider yourself lucky..

avatar

BlazePC

...

avatar

Whatme45

I always considered the life of a mechanical drive based on the how much the drive physically moves. Laptops and USB Drives (If it moves desks often) I would consider a high failure rate while desktops would be low.

avatar

John Pombrio

I have seen computers running testing and calibration equipment that were just beat to death with bumps, scratches, dents and footprints. No hard drives ever failed on them that I saw, heh.

avatar

TommM

Chalk me up as one of those who has never had a hard drive failure - and I'm 58, been building my own PC's for as long as I can remember.

I do turn my system off each night or if I'm not going to use for several hours, and also currently maintain 3 drives - an SSD as the boot/OS drive, a 1TB for programs and a 1TB for storage files. Also using an HAF X case with a brazillion fans, so everything's kept quite cool. I'm sure that distribution of files over 3 drives, and not keeping them running at all time has something to do with my non-failure rate.

avatar

wumpus

How quickly do you (and others with zero failures) retire drives? I'm 43 and started with an Atari 400, but I've lost a few retired drives and one primary drive with no warning (the failure was in the interface, and I eventually pulled all the data off of it). My father also lost a drive, but I doubt he would understand the warning signs.

The retired drives were when I wanted to test something on a spare computer, or dump some data on a third/fourth drive and discovered the drive was dead. I think I have one working IDE drive (was playing around with the Asterisk phone system), but most are dead.

avatar

MAIZE1951

I never experienced an hard drive failure, as I'm 62 and also been building my own gaming PC's (I love playing shooters like all the Battlefield series) for years also, as I got my start in building Heathkit electronic products. And yes you had to solder wires and electronic components as I built an nice Heathkit stereo and an Heathkit robot that still works.

avatar

The Mac

yup..i built the Heathkit robot as well in highschool. Man that thing was complicated.

I built the DMM, a stereo, a shortwave radio, and a couple of smaller ones as well.

Loved those things.

avatar

Peanut Fox

I've also found that using hairless fans helps a ton with cooling ;)

avatar

The Mac

Especially those hairless sexy Brazilian fans..

lol

avatar

USraging

+1

avatar

Gikero

I vaguely remember reading a article, that I thought I had seen on MaxPC, that Google posted about hard drives. It suggested that heat and drive failure didn't correlate.

"One of our key findings has been the lack of a consistent pattern of higher failure rates for higher temperature drives or for those drives at higher utilization levels."

http://static.googleusercontent.com/external_content/untrusted_dlcp/research.google.com/en/us/archive/disk_failures.pdf

avatar

devin3627

heat is a major error users make. businesses lock up their servers in airtight places without ventilation. always make back airflow with a pc. heat is number one, that circuit board frys, believe it or not... if hard drive chips were universally replaceable, the rates would go down!!!!!

avatar

BlazePC

Typically, server drives see much higher duty cycles across the gamut of functionality - hence why they need to be kept cool. Additionally, server drives tend to be higher generators of heat due to faster speeds - spindle motors and actuators.

Comparing server environments to that of low duty cycle (in comparison)
PC environments isn't applicable..

avatar

Obsidian

I have been building PCs for almost 20 years. I've witnessed my share of hard drive failures in that time. Back when IBM used to make the Diskstar drives, out of 10 drives, 5 of them failed within 1 to 3 years. A 50% failure rate. After 5 years not a single drive would work, click of death failures for 100% of them eventually.

Maxtor used to make some blazing fast spinning drives named Diamond Max Plus, or some such name, and I could never get one of those working for more than 60 days. They got so hot that the 80mm fans that were supposed to cool it, never quite kept up. I went through three return processes before eventually getting a refund.

Remember SCSI SCA drives? If you were trying to keep pace with Maximum PC performance benchmarks it was a tempting and expensive route to take. I put a few of these in a RAID-0 config and after 6 months had one of those drives catch fire -- melted the controller chips on the back of the PCB complete with black smoke and an odd looking blue flame. If I'm remembering correctly the drive was fittingly named the Quantum Fireball.

I consider the above situations to be poorly designed product issues and I just never purchased those brands again.

Ever since I have been using Western Digital drives and SSD boot drives I haven't experienced the infant mortality the article mentions, nor any outright drive failures. In a NAS device I did have a 3TB drive that shipped with bad sectors. After an RMA that was all fixed. I'm probably lucky for having so many successful drives after all this time.

I think the 26% failure rate might be accurate for spinning hard drives across the entire brand/type spectrum over the last 5 years. I don't build as many PCs as I used to, and I don't need as many HDs due to increasing capacity, so in the last 5 years I'd be a poor anecdotal metric for this comparison.

avatar

The Mac

oh geez, the IBM "Deathstar" drives...i remeber those...

avatar

John Pombrio

Me too. I had a couple go bad on me. One that kept working tho.

avatar

Hey.That_Dude

I've had two mobile drives fail on me. Click of death and a bad sector. The bad sector was after 2 months use, the click was after 3 years use. I've never had a desktop HDD fail, even after being struck with a "lightning strike" caused surge.
So this data sounds about right to me.

avatar

The Mac

im 46, been using computers since i was 13.

Ive also never had a failed drive, and i do not do anything special to mine.

Although i do tend to replace them every few years when i run out of space.

avatar

Xenite

I'm 35 now, been using PC's since I was 14. In all that time I've never had a HD take a dirt nap on me. I'm also one of those people who have never used the power button. :P

Have I really been THAT lucky? 26% fail rate in 4 years sounds insane to me. Granted I tend to pamper my HD's, I keep each one in a dedicated SYTRIN heatsink with it's own dual 40mm fans for cooling.