To emphasize how little data it collects, the NSA reported to the American people that it collects a mere 1.6% of the 1,826 petabytes of data that flow through the world's Internet pipes each day.
Tiny Numbers of Huge Files. One point six seemed like a small number unti I realized that the spooks were talking about the volume of data, not the volume of messages. Internet data volumes are huge due to the huge sizes of files being moved around. An HD movie is 5GB; Netflix is said to make up 40% of the US's Internet data volume, because its traffic consists primarily of huge movie files. But Netflix takes up a tiny proportion of total message traffic.
Huge Numbers of Tiny Messages. In contrast, messages take up tiny amounts of data. A typical formatted email is less than a megabyte; that's 5,120 email messages per single movie file. A text message takes 160 bytes. That's 26,200 messages per movie. It's mostly those movie, tor, music, and other large files that make up the 1,826 petabytes.
(1 petabyte = 1,024 terrabytes = 1,048,576 gigabytes = 1,073,741,824 megabytes.)
We see that the percentage of messages collected by NSA must be huge, much larger than the 1.6% claimed. They admit as much with their "connections of connections of connections" collection, meaning 100,000 to one million peripheral-persons per person of interest. This does not work out to a mere 1.6%.
This makes me wonder if NSA collects 98.4% of message volume.