Spinning hard drives have been around long enough that we have a pretty good idea of what to expect as it relates to their lifespan and even which ones are best to avoid. Because solid state drives are relatively new, there’s still a lot to learn as it pertains to flash-based storage.
A new report from researchers at Carnegie Mellon University should go a long way in helping us understand the emerging storage medium.
In it, researchers examined Facebook’s datacenters over the course of a four-year period and while it doesn’t specify the makes and models of drives examined, there’s still plenty to take away.
As The Tech Report notes, data was gathered straight from the drives’ hardware rather than an operating system, effectively measuring how much data was actually written to the flash cells. In the chart above, we can see that failures tend to happen early on followed by an extended period without any issues. This continues until the drives reach the end of their useful lives.
One interesting trend to note is that the initial use of new drives led the controller to immediately identify unreliable cells. Higher and lower capacity drives alike exhibit this behavior although the window for detection is shorter with smaller drives.
Operating temperature was also shown to impact reliability. Drives running at lower temperatures and those that used more aggressive throttling mechanisms exhibited fewer cell failures. Our own research on SSD drive temperatures can be found here.
The full report, titled “A Large-Scale Study of Flash Memory Failures in the Field,” is available online for those wishing to dig a bit deeper.
https://www.techspot.com/news/61090-researchers-publish-first-large-scale-field-ssd-reliability.html
