264 bytes = 16,777,216 terabytes
It’s actually not that much. It’s just 8½ million 2 TB disks.
Purchased at retail from Amazon.com @ $80 each, that’s $671 million, easily within the reach of governments and large corporations (a non-retail bulk purchase would be cheaper, even if you account for failed drives, adding some redundancy, and the disk manufacturers’ power-of-10 tax).
Do we need to think about 128 bit byte addresses for disks yet?
Richard WM Jones 
Regarding the “power-of-10-tax”: is there any reason to manufacture hard disks in capacities that align with powers of two? I don’t think so. We don’t use powers of two when we describe network bandwidths, so why should we do it at measuring mass storage space? Just because some ancient file managers did it that way?
Fortunately there are those nice little IEC binary prefixes we can use when referring to powers of two so we no longer have to abuse clearly defined SI prefixes. So yeah, one big btrfs filesystem is limited to 16 EiB. 😉
For SSDs, powers of 2 reflect how the hardware is organized (although not necessarily what a user will see).
Those IEC prefixes are ugly as hell, and who voted for / wanted them anyway?
haha, no.
imagine the size of the server room required to operate all of those?
think, cooling+power+staff+servers to actually even turn them on.
so actually, it is a lot.
It’s a certain number of data centres located around the world, but I doubt it is bigger than what Google or the NSA currently have.
I don’t think it’s an issue, because you won’t handle all that space as a single unit.
Instead you’ll be using something like OpenStack Object Storage, that can handle more than that (theoretically!) and a 64bits cluster will be fine.
Although thinking about it, 3 replicas of each byte sound like a total waste with 16 millions of terabytes! 😀
Pingback: Partitioning a 7 exabyte disk | Richard WM Jones