Don't Yell at Your Hard Drive!

No, you won't hurt its feelings. But as this video shows, hard drives are susceptible to vibrations.

Yelling, or more likely other loud noises, will adversely affect the drive. So, no jackhammers, explosions, etc. <grin>

My question is why the other steady state loud noises in the data center did not cause problems?

I doubt very much the vibrations in his shouting affected the disk in such a dramatic way. There is a huge impedance mismatch between air and the materials in a disk drive, so the energy delivered to the drive is minimal, with nothing like the energy of vibrations delivered to drives by earthquakes or even a heavy truck driving nearby. Humans produce 100% humidity in our outgoing breath, and it is conceivable that the humid air he blew at the drive could affect a disk drive, particularly if the drive array was set up with sensors to slow activity when there is extraordinarily high humidity.

Fun post, though.

I never, ever yell at my hard disks.

I do however often yell, even scream, at Photoshop (but I love it still).

I have never seen any reaction to the yells, except from my dog, who gets very excited...

No hard drives were injured in the making of this post.

@ Hector Javkin: Then there is this story in SANS NewsBites Vol:20 Num: 033

"Loud Noise Damages Nasdaq Servers in Sweden

(April 26, 2018)

"Noise from an activated fire suppression system caused Nasdaq servers at a data center in Sweden to fail, shutting down stock market trading in Sweden, Finland, Denmark, Lithuania, Latvia, and Estonia, as well as other markets in Iceland. Data centers often use systems that replace oxygen with inert gas to suppress fires; the noise from these systems can sometimes be extremely loud.

"Editor's Note

"[Ullrich]

This is an interesting and often forgotten phenomenon. Unsurprisingly, mechanical hard drives are very sensitive to vibration. Loud noises, in particular at the right frequency, are able to sufficiently disrupt hard drives to force them into "auto protect" mode which disables them temporarily, or to permanently damage them."

Brooks, I'd be interested in what the intensity and spectrum of the noise from the fire suppression system was. It's possible that a fire suppression system might produce high intensity at frequencies that resonated with something in the computers that were affected. The spectrum of a male human shouting is rather diffuse, so there isn't a great deal of energy at any particular frequency. I would be less surprised if the human voice could damage a computer by singing, which often yields a very narrow spectrum. When that spectrum matches the resonating frequencies of glass, for example, or perhaps some component in a disk drive, it can cause damage.

Best regards, Hector

And another urban myth is born!

'Slow news days' should never be mistaken for evidence or proof of anything.

Rodeo_joe, SANS is a well respected security organization.

SANS Institute: About

They seldom publish "myths" urban or otherwise.

Hi Hector,

If you watch the video in my original post, the yelling did not do permanent damage to the system or cause a crash. It did show up on instrumentation that showed an increase in latency on the disk. I suspect you would damage your vocal cords well before you did any real damage to the disks. In fact the video was a promotion for the monitoring software as well as being a fun, interesting video - well fun if you are a "geek".

On the other hand the (I assume) much louder noise (and perhaps the gas turbulence) of the fire suppression system, did cause a system outage attributed to disk vibration.

The yelling is mechanical energy that causes the HD case to vibrate and this vibration gets transferred to the actuators that read/write information onto the HD platters. There are sensors on the actuators that detect the unwanted motion and data reads and writes are supposed to be suspended vs risk corruption. All kinds of mystery patents on how this works.

This was a big problem in the early 2000s to teens as HD density rapidly increased but design tolerances didn't until recently.

What didn't help was the Satan known as RAID 5 was still common in SAN setups. Essentially motor vibration from dozens of 10k drives would resonate in some enclosures and usually knock out the top tier of drives because vibrations were strongest there, and RAID5 wouldn't tolerate dual faults at once. Bang....data volume gone, and mysteriously in pairs with RAID 5.

Drive makers have claimed newer SAS drives have better dampening for this problem. With more companies moving to cloud and SSD and local storage taking over smaller on prem server duties it's far less of a problem anyways. I still cringe when I hear EMC and RAID 5 in the same sentence.

"Rodeo_joe, SANS is a well respected security organization"

- It may well be, but the reporting/investigation of the incident may not be accurate, thorough or authoritative.

Why was the fire suppression system activated in the first place? These things generally don't 'go off' for no reason. Maybe the cause of the fire suppression activation was also responsible for any HD data loss? Maybe there was a glitch in the power supply? I doubt that a completely forensic examination of the lead up to the incident was carried out. And if human-error was involved, then someone would have a vested interest in diverting blame.

In short, it still seems like a case of sloppy reporting and conclusion-jumping to me.

Sure, I'm not going to deliberately thump the desk that my portable HD sits on. OTOH, I'm not going to tiptoe around it and speak in lowered tones next to it either.

I'm pretty sure that human ears would suffer damage before an acoustic dB level high enough to disrupt an HD was reached.

Essentially motor vibration from dozens of 10k drives would resonate in some enclosures and usually knock out the top tier of drives because vibrations were strongest there

Perhaps I'm weird in this, but I've found every 10K drive I've used to be obnoxiously loud. Most of my experience is with Seagate drives(I think Cheetahs) but I've had Hitachi and a few other brands in 10K. They sound like a chain saw to me. That gets amplified in the enclosure.

To be fair, though, I think the SGI Octane that runs in my office all the time has 3 10K drives, and although it's loud it's nothing like others I've used.

15K is a different story-while the ones I've used(and yes, that includes some 15K Seagates) do make some noise they sound more like a normal 5400 or 7200 rpm HDD.

Most 15k drives I've worked with don't live long enough in high production environments to notice :-) Twice as fast, twice the heat...twice the wear and tear. SSD's to the rescue.

Fun pulling them out of hot swaps though when still running and feeling the torque.

Fun pulling them out of hot swaps though when still running and feeling the torque.

Even at 7200, the gyroscope effect is fun. It's REALLY pronounced in 15K drives.

My excited dog after yelling

Perhaps I'm weird in this, but I've found every 10K drive I've used to be obnoxiously loud. Most of my experience is with Seagate drives(I think Cheetahs) but I've had Hitachi and a few other brands in 10K. They sound like a chain saw to me. That gets amplified in the enclosure.

 

To be fair, though, I think the SGI Octane that runs in my office all the time has 3 10K drives, and although it's loud it's nothing like others I've used.

 

15K is a different story-while the ones I've used(and yes, that includes some 15K Seagates) do make some noise they sound more like a normal 5400 or 7200 rpm HDD.

That is why the HIGH RPM drives are in server rooms, not desktops.

Couple that with the 1U servers, with NOISY small fans, and ouch my ear.

A friend of mine said that is why even though he has access to 1U servers, he will never have one in his house. TOO NOISY.

The yelling is mechanical energy that causes the HD case to vibrate . . .

Didn't see this thread for a while. Yelling at 120 dB (which is extremely loud yelling) yields about 0.0001 watts in air, which (because of the impedance mismatch I mentioned earlier) is very inefficiently turned into vibrations on the case of the drive.