A cloud backup service has, more or less, the same needs we end users have: they have to buy hard disks, just like us (even if in far greater numbers); those drives must work at all times: if they go kaput, thousands of users will lose data, just like it would happen to us.
The scale is obviously incredibly wider, compared to a single user, thus Backblaze watch their server’s performances closely, and they annually publish a report with malfunctions and shutdowns figures, ordered by model. In 2016, only three kinds of hard disk had no issue whatsoever: an 8TB HGST model, an equivalent Seagate model, and a 4TB Toshiba disk.
We suggest you read their original post, rich in useful information. Like many solid companies, Backblaze’s transparency is refreshing, and a behind-the-scene look is always intriguing.