Failure Prognosis Engine for HDDs in Notebook PCs

With users' approval, Toshiba has been collecting operating logs from more than 1.66 million notebook PCs around the world via the Internet. We analyzed the relationships between these logs and repair data by means of a data mining technique to develop an HDD failure prognosis engine. As a first step, we defined more than 700 characteristic variables for time-series data that capture peculiarities of a failing HDD. In creating the engine, these variables were combined in an optimal manner to achieve the maximum accuracy.

It was verified that the HDD group alerted by this engine had 67 times higher failure probability compared with the non-alerted HDD group, indicating that PC users can prevent data loss if they are aware of a sign of a failing HDD.

The HDD failure prognosis engine is being incorporated into the management tool of our notebook PCs and will be available in 2013.

Failure prognosis engine for HDDs using data mining analysis