hdd-lifetime-prediction

A quick ML inference engine to predict the remaining life of a hard disk drive.

This is an implementation of the model described by the Amram et al. (2021). The study used Backblaze data to train an predictive model that estimates the remaining useful life of a hard drive.

Usage

A basic example of this tool can be shown below:

docker build -t hdd-lifetime-prediction-app .
docker run -p 8080:8080 hdd-lifetime-prediction-app

sudo smartctl -A /dev/sdX | curl -X POST -H "Content-Type: text/plain" --data-binary @- localhost:8080/hdd-lifetime-prediction/

Output:

{
  "predicted_lifetime": 276.3315409152338,
  "predicted_stats": {
    "expected_lifetime": 276.3315409152338,
    "lower_lifetime": 21.086563249558207,
    "lower_node": null,
    "median_lifetime": 263.7432142532636,
    "n_samples": 6167,
    "split_feature": null,
    "split_threshold": 0.0,
    "upper_lifetime": 608.2135071357179,
    "upper_node": null
  }
}