How sensitive are they? Well, I have seen some overblown claims from SSD makers that shock will cause HDD head crashes. I am not sure that I believe such claims, but I certainly do believe that an HDD’s actuator (the read/write head mechanism) can be shaken away from its track, causing a one-rotation delay before it can regain its position and access the desired sector. After all, on a terabyte HDD the tracks are only about 100nm apart (that’s one ten-millionth of a meter)! It doesn’t take much vibration to move the head more than that much, so disk array designers pay a lot of attention to shock isolation to make sure that an actuator movement from one HDD is less likely to vibrate an adjacent HDD. Even fans are a concern.
The graphic for this post is from a video that was made in Sun Microsystems’ Fishworks Lab way back in 2008. It is only 2 minutes long and is well worth watching. The engineer in the photo, Brendan Gregg, shouts his lungs out at a disk array, and the video shows a significant latency spike as a result.
Notice how loud the data center is to begin with, and you will wonder why vibration isn’t more of a problem even without the shouting.
In the data center this may not pose significant problems, especially if you don’t make a habit of shouting at the disk arrays, but in other environments, especially in industrial applications, this exercise makes it clear why many engineers choose to use an SSD rather than an HDD, even if they don’t need the the SSD’s speed. I like to give the example of a jet fighter that shakes so much that the pilot may even worry about losing the fillings in his teeth.
(Thanks go to Tom Coughlin of Coughlin Associates for his help with the track widths, which he put as 300,000 tracks per inch. Wow!)