What we found when we simulated the backbone of the Internet

It sounds like they found an advertisement for SpatialOS.

A single BGP routing table does not. A full table, multiplied by the number of unique autonomous systems however, can quickly reach that size.

not that i recall, but in the 90s, construction crews would routinely cut backhaul fiber lines and large areas of the country would go offline for most of a day while they fixed it, literally by stitching the cables back together. i think that's when they started putting markers by all the cables, because it was becoming a real problem. "call before you dig!"

or, more commonly, entire sites would be unavailable for a day or two, because everyone ran their servers on insanely expensive and unreliable T1 (or even crappier leased line) connections that were also prone to getting cut. that's when the colo / datacenter / hosting business really took off, because imagine running something that makes you money and relying on your local shitbox CLEC telco to deliver it for you. untenable at best.

for example, i remember a guy who ran a small gaming server outfit on a set of 56k leased lines running out of his garage. if you ran a clan, you could pay him for low-latency practice space. that was considered a legitimate internet business, cutting edge and visionary, even. i think it was a side business, and took off when the industry matured a little more and he could actually deliver decent solutions at consumer prices.

aside from basic nostalgia, those were not the good ol days. in fact, it pretty much sucked. flakey hardware, buggy software, slow-as-shit networking, no wifi ... the struggle was real.

these days, everything that comes out blows your fucking mind at how fast, reliable, high quality, and cheap it is. imagine waiting years for an incremental improvement in bandwidth instead of ridiculous shit like google fiber replacing your good-enough 100 megabit cable ISP. it's much better now in absolute terms, no matter how much people complain. and i'm only in my early 30s, i got an early start, but to someone in their 50s it must be mind-bending how much progress has been made.

The 1998 Morris Worm was as close I as I can remember it getting.

Sorta related: "All Circuits are Busy Now: The 1990 AT&T Long Distance Network Collapse": http://users.csc.calpoly.edu/~jdalbey/SWE/Papers/att_collaps...

https://en.wikipedia.org/wiki/SQL_Slammer (2003) was pretty close.

TAT-14 going down was the closest thing to this in my lifetime. It was a pretty crazy time. But it affected the UK more than it did the US.

The Arpanet outage of Oct 27, 1980 pretty much took the whole thing down (slightly before the name changed to Internet). See RFC 789. A single flakey IMP announced it was zero hops from all others, so all traffic went to it.

The Internet was funded so this couldnt happen. The government asked AT&T to built a fault tolerant network, and AT&T didn't see the short term profitability or point or think it was possible. They didnt even bid the job. The government was worried that major cities could be taken off the map, disconnecting the coasts. I would say the closest thing to the entire internet crashing was 1969.

1964 - Paul Baran published On Distributed Communications: Introduction to Distributed Communications Network, which discussed packet switching networks. It also outlined the concept of fault tolerance, or the fact that there is no one computer responsible for the entirety of the internet. If one computer fails, the internet continues. https://www.rand.org/content/dam/rand/pubs/research_memorand...

1968 - Lawrence Roberts and DARPA came to a final plan and set of specifications for ARPANET, and released a request for quotation, that BBN won, which was basically a contract to produce the routers. It should be noted that larger companies including ATT did not even bid on the contract, saying such a network was impossible.

1969 - Four nodes were added to ARAPANET, including UCLA, Stanford Research Institute, University of California at Santa Barbara, and the University of UTAH. It should be noted that the first connection, between UCLA and Stanford, caused the Stanford computer to crash after typing l and o of login