Dragoneer, on reddit, is blaming old hardware:
To his credit, he at least understands half of the problem that he's trying to solve here. He's pretty strapped for resources. That's a fact.
Wanting newer hardware for more rapid backups, though, is a bit misguided. Really, the thing that he should be optimizing for is rapid restores. I don't give a shit how quickly you can back up data because what I ultimately care about is being able to restore service, in a timely manner, to where it was before disaster struck.
For those playing along at home: If your storage hardware is slow (hint: it's slow), find a good way to take deltas of your database, and keep a rolling log of files that get written to disk via user uploads. Write that to your backup media. Make your application and systems aware of this by both notifying them when and what things have been backed up and making them consume those notifications.
Then, after you've backed stuff up, periodically try to restore a random sample of that data. You know, just to make sure that your backups actually work.
Why we call it a backup system is beyond me; it should be called a fucking restore system. Anyway, RPO, RTO, and MTTR, motherfucker.