I had an interesting conversation this weekend with a friend
who handles what use to be Big Data but these days is more like Moderate Sized Data.
He's got a couple of interesting problems at the moment.
1. Somebody parceled out a $300 million budget, but forgot to set aside the $15,000 they need to upgrade a data array from 1TB drives to 4TB drives.
2. He sometimes has to "archive" data that is old and not being actively used, but which has been distributed to external users (3+ years old in most instances, some of it as old as 10 years). In the past he's attempted to use external USB devices. His experience is that whenever he writes 2TB of data in one go, the external drives die.
3. They once had a tape drive from Sun, but for whatever reason, it never worked with their server setup. They let their support agreements expire when Oracle acquired Sun and migrated to Linux. After thinking about the storage problem, he thinks it might be time to investigate getting a tape backup unit again.
Personally, I never deal with this much data. But based on his experience, I think you need tape in the system somewhere.