I pinged every IP address that wasn’t reserved. The image is 8k by 8k and is re-encoded as an AVIF to be friendlier to mobile devices. Like every other survey done, it is using a Hilbert Curve to convert the linear address space to a contiguous 2d space. The hotter the colors (blue is coolest), the denser the ping responses were.
(If you are interested the full-resolution pyramidal-tiled TIFF can be downloaded and viewed in QuPath on desktop. I’ve also compressed the ping response data into its own format down to about 150 MB. PM me for a link)

Here is a 2006 survey to compare.

Some observations: Big Tech (USA) is in the top left. US government allocations, for the most part, did not respond to any pings. And maybe you didn’t realize this before, but Multicast (Class D) & Class E consume a whopping 12% of the IPv4 range.
every IP address
every IPv4 address
ipv6 would look like a starless sky
IPv6: a dark image
I left details out to be less verbose.
Removed by mod
about a month, since I ran it at about 6000 addresses / sec checking distinct addresses 4 times (round robin) or 1500 finished / sec. This was the fastest I went to avoid silent UDP drops and to hopefully not annoy my VPS hoster too much.
How long did it take to ping the whole world?
It would be nice to have some overlay of the two images with a difference to see what changed.
Forgive me if this is an ignorant question. How did you do the incremental address to address search? In my head I would start with “000.000.000.000” and then “000.000.000.001” and so on but that doesn’t account for the “000.000.000.1” scenario, or any combination thereof.
Writing it out like 000 is just a convention.
.1is the same as.001. They are actually hex numbers from0toFF.Ahh…makes sense. Thanks.
Essentially, IPv4 addresses[2] are just numbers from 1 to 2147483647, for example 3405804031. However, since address routing is often based on common binary prefix, more intuitive methods like 203.0.113.255 are used. This notation is just a length-4 list of integers from 0 to 255. To convert from an integer to a common IP address, you divide-with-remainder with a constant divisor of 256. For example, 3405804031 /% 256 = (13303921,256), 13303921 /% 256 = (51968,113), 51968 /% 256 = (203,0), and 203 /% 256 = (0,203).[1] Collecting all the remainders in reverse order, then joining them with periods, produces 203.0.113.255. (This notation is just for people to read; aside from parsing code, none of IPv4 uses this notation.) When you enter an IPv4 address, the opposite happens — an expression like ((203 * 256 + 0) * 256 + 113) * 256 + 255, which evaluates to 3405804031, is performed. These octets are just numbers — using 203.000.113.255 or 203.00.113.255 in place of 203.0.113.255 is merely a choice of how to write the address, as 203 * 256 + 0 = 203 * 256 + 000. mraow
[1]: this operation can be omitted; I include it for symmetry. [2]: though all of this holds true for IPv6, I really didn’t feel like going through 128 bits of address
Thanks! And thx to all.
Where is 100.64.0.0/10. Should be reserved
Unlabeled grey square at the bottom
Exactly the same thing had been done in this very interesting video I had watched some time ago. Did you get the idea from there?
Yeah, a link to the tiff would be great.
Updated
Thanks!
deleted by creator
The Hilbert curve visualization is surprisingly satisfying.







