Calculating storage requirements for UTXO

[indolering]

From Aarron Schwartz's [Square Triangle FAQ](https://squaretriangle.jottit.com/faq),

Total size doesn't seem to actually be a problem, surprisingly. We have 1.9e8 domains right now; if we suppose each name is 20 bytes long, has a 20 byte nonce, a 20 byte revocation key, an 8-bit timestamp, and maps to an IPv6 address, we're looking at a per-node total database size of about 16GB --- big, but not unthinkable.

20 bytes * 3 + 8 bits = 61 bytes. However, this doesn't line up with the numbers above.

Code: Select all

domains * bytes = totalBytes/bytesPerGig = totalGigs

which is

Code: Select all

190000000 (domains * 61 bytes) / 1073741824 bytesPerGig = 10.8 Gb

Working backwards from Colascione's 16GB estimation should be

Code: Select all

totalGigs * bytesPerGig / domains

or

Code: Select all

16GB * 1073741824 bytes / 190000000 domains

which is ~ 90 bytes per domain.

Am I just fucking up the algebra or is something else wrong here?

[midnightmagic]

By *his* algebra we have:

20 bytes(name) + 20 bytes(nonce) + 20 bytes(revoke) + 1 bytes(stamp) + 16 bytes(IPv6) = 77 bytes

77 * 1.9e8 = 14630000000 bytes, or 13.625GB for the *raw data*. He's estimating up for database overhead perhaps, or he was just doing the math kind of in his head and coming up with a raw figure.

He's right, it's not that much, not really. The magnitude is really the important thing anyway.

[indolering]

By *his* algebra we have:

20 bytes(name) + 20 bytes(nonce) + 20 bytes(revoke) + 1 bytes(stamp) + 16 bytes(IPv6) = 77 bytes

77 * 1.9e8 = 14630000000 bytes, or 13.625GB for the *raw data*. He's estimating up for database overhead perhaps, or he was just doing the math kind of in his head and coming up with a raw figure.

He's right, it's not that much, not really. The magnitude is really the important thing anyway.

Duh, thanks.

[indolering]

One more question, would the UXTO data structure require holding the entire /d space or can I choose a subset? The Zipf distribution would enable skimming the to 20% most important sites in client side storage.

[indolering]

From IRC:

It depends on how they do it. They could verify back just to merkel root of the generation tx.