An excellent explanation of hardware failures is provided in the section for data recovery.

These events may happen once or several times until the hardware failure occurs, or not at all.

They include hardware failures, software issues, or people and process issues.

Drives can be designated for use by an array in the event of hardware failure.

Now you can lose 4TB of data all in one hardware failure!

This also means that the nodes themselves are less likely to have hardware failures than fat clients.

What trends are we seeing for hardware failure and life expectancy?

An Amazon official later stated that it was due to a hardware failure.

It creates a single choke point where a hardware failure will take down the entire system.

It's just another option when what they have can no longer function (usually because of old hardware failure).