Correlation Between Hard Drive Power Cycles and Failure

[Jared]

Here's another interesting stat from Backblaze regarding the correlation between the number of times a drive is power cycled and it's failure rate:

View attachment Number of power cycles.png

Obviously there will be some correlation between age and number of power cycles. However the chart here is far more pronounced than the hours of operation charts are. So it seems to back up what I've always said, it's better to leave the drive spinning and disable the silly power saving modes that turn off the drives.

 

[Jared]

I would guess that the dip in the chart around the 52-65 raw value is about the time when all the Seagates had failed and only WD and HGST drives remained. Haha.

 

[LarrySabo]

The chart only goes to 91 power cycles, so is meaningless, as far as I can see.What am I missing here?

 

[Jared]

The chart only goes to 91 power cycles, so is meaningless, as far as I can see.What am I missing here?

That's not the number of power cycles, it's just the RAW figure pulled from the SMART data. So each number might be more like 1000 power cycles.

 

[LarrySabo]

I've always interpreted Raw values as absolute value for the attribute, and Current or Normalized Value as what they are calling the Raw value in the graph. Guess I should read the article. Also, the further Current/Normalized Value is below the reference value (often 100 or 200), the worse off the drive with respect to that attribute. Yet in the graph, it suggests the higher the value, the worse off the drive.

 

[Jared]

I would assume they reversed the figure to show the drop in RAW value so the chart would make more sense visually.

 

[LarrySabo]

I've just read the article section that deals with power cycle count (S.M.A.R.T. 12), and they are indeed using the raw value, not nomalized value (which was 100 for all the drives in their sample, in spite of varied raw values). No mention of reversing raw values for the graph. They do mention that their drives have been power cycled less than 100 times, hence the highest raw value of less than 100. While the correlation is strong, they do say that power cycles may not be the cause of the failures.

 

[Jared]

Yes I see you are right about that. Their pod system counts the number of actual power cycles. I guess with less than 100 the SMART figure wouldn't have time to move.

Given that the drives in my systems are almost never power cycled though, I'd imagine my drives are likely under 100 as well. Might explain why my drives never seem to die.