"""
For commonly used 1024-bit keys, it would take about a year and cost a "few hundred million dollars" to crack just one of the extremely large prime numbers that form the starting point of a Diffie-Hellman negotiation. But it turns out that only a few primes are commonly used, putting the price well within the NSA's $11 billion-per-year budget dedicated to "groundbreaking cryptanalytic capabilities."
"""
My understanding is that they record encrypted traffic too. They can't read any of it - yet.
But they're betting one day either a security vulnerability will be discovered, or computers will be fast enough to attack the encryption and allow them to read the data. So even though it's unreadable today, it might be in 10 years.
Even 5EB would be a stretch for 2013. 5ZB is flat-out impossible. As another poster points out, that's years' worth of total worldwide drive shipments (most sources put it at less than 1ZB in 2013). Large buyers are further constrained by the fact that their demand can cause price spikes even at much lower percentages of the total. Not even No Such Agency has that kind of budget. The Utah facility also isn't physically big enough for that figure to hold. I work on large storage systems at one of those large buyers, and I've toured one of the several data centers where ours live. NSA's Utah data center looks to be on approximately the same scale, not orders of magnitude bigger. It's further plagued by power problems, which is another constraint on total size.
So I looked into that quote from the NSA director. What was actually said, apparently, was that the center was designed to hold up to 5ZB, not that it actually did. That seems to be a design based on some extremely optimistic assumptions about future drive density, power consumption, and cost. Needless to say, those assumptions were a bit silly at the time and have only seemed more so in retrospect.
For platter or SSD drives, sure. Some forms of magnetic tape storage can get up to 300 TB per cartridge though, which can scale up to petabytes in the right config.
Still ridiculous for information that is worth less and less over time.
300TB per cartridge in 2013? I think 10TB per cartridge was pretty high round then. That's still around half a billion tape cartridges, which are also very slow to read and write from.
Do you have a source for that 100 million figure? It strikes me as awfully low considering how many personal computers must surely exist in the world and how often they'd be replaced without reusing the drives, not even factoring in servers.
I remember hearing a speech by a quantum computing researcher that was primarily funded by the NSA. He included an anecdote about how “they would prefer quantum computing didn’t exist, but if it’s going to exist - they want the first one”.
"SSL added and removed here! :)"