(C) Alec Muffett's DropSafe blog.
Author Name: Alec Muffett
This story was originally published on allecmuffett.com. [1]
License: CC-BY-SA 3.0.[2]


Notes on mining the #facebookcorewwwi onion address

2022-03-01 23:45:48+00:00

Seeing as it has been conveniently leaked by Frances Haugen, I thought it would be nice to write briefly on the mining of the (now defunct) Version-2 Tor Onion address which was known as “facebookcorewwwi” — the v2 onion-address for Facebook.

As already reported — because we shared that much — the onion address was a fluke, “mined” in a way similar to bitcoin, creating random keys until we found ones that we liked; and the ones that we chose to like were those addresses beginning with 40 bits that spell “facebook” when rendered in base-32.

Put differently: on average you will expect to “like” one onion address for every 240 random addresses that you create; or in other words you will “like” one onion address in every 1,099,511,627,776 — i.e. about 1 “like” per 1.1 trillion.

What we never got around to sharing publicly was how many candidates we mined; as my Haugen-leaked “Goodbye” post mentions:

[Colleague] wearing shades and drinking beer at 2am whilst herding 24 thousand onion-miners… https://alecmuffett.com/article/15058

…which is absolutely true: by chance there were a lot of machines being commissioned for a new compute farm, and we went to the team responsible and offered “onion mining” as a burn-in test for the hardware. This also meant writing custom code and also shepherding 24,000 machines to ensure that they were running at a representative peak performance for the test period.

They ran for an average of about 7 days, and in that time it produced a bit less than 250,000 of the “facebook”-prefix onion addresses that we wanted; from this we can compute some rough and ready estimates:

Reader Beware! Faux precision — I can’t be bothered to round these numbers up or down in a scientific manner, and we start from a rough memory/estimate…

candidates: 250,000 (ballpark, probably a few thousand less)

raw onion rate: 1 in 2 40 = 1 per 1.1 trillion

= 1 per 1.1 trillion raw onions generated: 274,877,906,944,000,000 (274 quadrillion, ballpark)

seconds per week: 604,800

onions per second: 454,493,893,756 (ballpark)

onions per second per CPU: 18,937,245 (ballpark)

candidates per second: ~2.42 (ballpark)

doing 240 work every (1/2.42) = 0.41 seconds (…)

Speaking of environmental impact — some kind of test/burn-in for a new datacentre is standard procedure, so we piggybacked on the new hardware to obtain an eco-free ride; we made a couple of stabs at estimating the power cost of mining those 250,000 candidates, and our semi-educated guesstimates were in the range of $5k to $15k — which sounds a lot, but you can pay much more for a top-tier “.com” domain name.

It was an excellent onion address, and I am glad we found it; and I will never forget the moment — as ever, the best ideas hit you in the shower — when I worked out that it could backronym into “facebook core www infrastructure“, which sealed the choice.

It helped a lot to normalise onion services, and I hope to see a lot more of them in future.
[END]

[1] URL: https://alecmuffett.com/article/15996
[2] URL: https://creativecommons.org/licenses/by-sa/3.0/

DropSafe Blog via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/alecmuffett/