Any time all your data’s on the same drive in the same place, you should be concerned. Even the worst drives have fairly low AFR, but every model is subject to bad batches or correlated failure. All four of your drives are probably from the same batch, same age, subjected to the same conditions. Odds are high they’ll fail around the same time.

Thanks. Appreciate the shares.

The primary reasons we use our current method are 1) because it is easy to calculate and understand, and 2) it works for us for the moment. The second point has to do with the cost of the components to buy and replace them. Replacing a hard drive is a trivial cost for us, failures are low, and substitutes are readily available. Meaning there is little incentive to make the forecasting process more precise. That will most likely change over time and scale, but for now… All that said, a few folks have played around with other methods like Kaplan Meier, so maybe we will as well.