Traffic on a Dead Net
05Dec06
Bockwurst and I were discussing an interesting question today: how much Internet traffic would remain if every human on Earth suddenly died? The Internet was designed to withstand a nuclear attack, but what would it be like if it did? What kind of applications would still be exchanging data with one another. How much data? How long would the power last? One imagines email clients and RSS readers would continue fetching on desktop machines, operating systems and security packages would continue phoning home for updtates, DNS servers would continue to exchange names. What other services would be running? Anyone care to offer a back-of-the-napkin calculation of how much traffic would remain, and for how long?
cool question. mechanical turk would officially be dead. heh. i’ll need to think about this s’more… right now i’m preoccupied with the origin of that image. seems like a ken-ichi original.
well, it would depend on who started the nuclear attack, if it was the cylons, then I would guess a lot of traffic would still be on the network!
Most of the remaining traffic if we got rid of all the Myspacers and home shoppers and bloggers would be coming from sensors of various kinds that would continue to “call home.” These use very little power and thus would operate for years. There are an enormous number of these already deployed, ranging from the mundane RFID tags that report their location to more interesting devices that remotely monitor equipment like oil wells or pipelines. Some of these are even solar-powered, so these could continue to feed their little data packets into the network forever.
And this reminds me that the roof on part of my house in St Helena is covered with solar panels, and they report their power generation into the network so that I can remotely monitor my electricity production and consumption. I suppose that it would be possible to set things up so that the solar panels would generate the power to reboot the modem etc and restart the traffic every morning when the sun rises. The solar panels are supposed to last for 40 years…
some random apparently relevant links:
http://www.sics.se/~adam/ewsn2004.pdf
http://www.sics.se/~thiemo/lcn2003.pdf
bob glushko
Kevin, I’m so happy that someone mentioned The Turk! A lot of us were interested in mechanical and puppet theater in my program at college, and the performance aspects of it. I’ll try and did up relevant portion’s of my advisor’s thesis (possibly now book?) on The Turk and give it to the local oaf.
it would be my guess that even though there would be a lot of machinery wanting to use the net, the net itself would go down pretty quickly, because of some unforeseen problem with routing or name serving or traffic policing. one of my favorite thing i ran into recently was the notion of that on he one hand, for robust systems to exist, engineers must build them to withstand unforeseen problems, while on the other hand, by the very definition of “unforeseen”, this is not possible. so in the end, you always need people being smart enough to cope with emergencies, and no matter how great the net works today, i think that without some really smart people monitoring it, it would go down pretty quickly. i have no proof for this whatsoever, but i do think this is what would happen.
however, if we assume that the net would remain available as a transport infrastructure, then we would have have the same problems with our computers, most of them going down pretty quickly. after the latest ie7 patches to my glorious windows, for example, my laptop is back to the good old days to at least one crash a day. so without me rebooting it, it will also die very soon.
however, if we assume that the computers using the net would remain available, then we would have the sensors and time-triggered things still running. i have not seen any recent traffic analyses of the net, but i would assume that it is still a lot of crawling, file sharing, and spamming. file sharing would go away pretty quickly, but crawling and spamming would probably survive a bit longer. but even they will probably not run very long without human intervention, they are automated processes, but they are also so complex that they need humans to control the overall process.
and then we have the “semantic web” and “agents” things, which have been touted so loudly by people in search of research money. maybe we would have some of these autonomous agents still roaming the web, searching for the lowest fare to oahu *and* and a good deal on a rental car there. but i don’t think there is a lot out there that is really autonomous in the strong sense of the word, most of that is simply crawling followed by some data mining. my guess is that this kind of activity would go away as quickly as search.
in essence, i think that the equation ? – ? = ? would yield a very small fraction of what we see today, at least after a couple of days. and pretty quickly, the net would collapse anyway, leaving the web without an infrastructure to run on…
Those remaining few bots will continue to win free iPods by smacking on flash banners of Bush. USGS’ installations will keep looking for those 2.5-ish tremors just in case N.Korea survives to perform nucluar tests. Anti-virus softwares and Windows will make the best of it by incessant downloads of updates and upgrades. Google will probably stop running soon after the blast not withstanding the enormous power hungriness of its server farm. The webcam on top of Sproul Plaza will continue to webcast the ashes in front of the MLK building. Phones will keep ringing to entice with the new offer to switch from AT&T to Comcast and Bangalore will be waiting to be paid for having processed some of this data.
wow, kesava, i never knew you had such a dark (great) sense of humor
i remember reading Nevil Shute’s On the Beach in grade school. it talked at length about how Australian animals such as kangaroos and rabbits would have the best chance of surviving a worldwide nuclear incident.
I wonder if any animals will continue to use/manipulate the information infrastructure after humanity ceases to exist. maybe pigeonblogging?
all this talk of nuclear winter reminds me of a Homer moment, after Springfield gets nuked out:
% Homer drives around in the oddly skied Springfield. He comes to a small
% traffic jam. Impatiently, he honks his horn.
It’s green, moron! Earth to stupid guy. Hello?! (he gets out and strolls to
the car in the front of the queue) Maybe a little friendly punching will
move your tail! (he punches the unskinned skull of Kirk Van Houten, which
crumbles under his fist and drops to the ground) Still got it! (claps the
dust from his hands)
– Homer, Treehouse of Horror VIII
devices that are solar/wind-powered will last until the next cylon attack; devices on an electric grid situated in a minority of organized countries will last 2 billing periods, then a trigger for bill-not-paid will cut off the electricity (I wonder, does Google pay its electricity bills with a cheque, or direct billing?); in the rest of the world, unpaid bills can accumulate for years before any action is taken, so it won’t matter. So, the post-nuclear internet will be environmentally sustainable, anarchic and will speak in tongues…
k7: I searched and searched for an image from that very episode, but, alas, failed, and so drew my own.
Bob & Dret: think I have to side with Dret on this one, for the most part. Computer systems are so very brittle, so I doubt many of them would last very long. Not sure how robust our electricity infrastructure is, but given that power outages seem to happen with human administration on monthly basis, and full on rolling brown outs have happened at least twice in the last 6 years or so (CA, NYC), I imagine only extremely durable systems would last more than a few months. Bob, even your roof would probably kick it as soon as the software locks up or a tree falls on it, and I’d put that between months (software) to a decade (tree).
Elisa & N8: Cylon attacks require Cylons. I put the rise of hostile AI between 100 and 300 years from now.
All that said, there is a burial mound in Ireland called Newgrange, where for 5000 years a narrow shaft of sunlight has illuminated the inner chamber only on the Winter Solstice. Although I can’t seem to corroborate the legend, I recall hearing speculation that the intent was to allow the spirits of the dead to return to Earth, if only briefly, to revisit their past lives. Reliable soul transportation, Mean Time To Failure 5000 years and counting. Prehistoric Celts: 1. Software Engineers: 0.
elisa: the “next” cylon attack? does finals week make you feel like cylons have already attacked?
my honest answer would be that i would not expect any network to remain running for very long, maybe several months for less vulnerable parts of the network, probably days, to hours for specific locations. in the interm, machines like crawlers, backup processes, content aggregators and spam bots would keep doing their thing, and the nsa’s comps would probably still be preforming some elaborate nlp to glean any information about the next al queda attack. as electricity became scarce, backup generators would kick in, and keep things going until the diesel they were burning ran out. slowly portions of the network would drop off, increasing load on other parts, which in turn might strain those parts of the network until they eventually broke down as well. finally, without the massive infrastructure we have abstracted away into monthly PG&E and (insert phone/data/tv provider here) bills, nature would quickly subsume most of our refined resources and civilization.
would google “notice” that we’re all dead?
the crawlers must increase or decrease the crawl frequency depending on how often it notices updates.
how would it react to having no ad clickthroughs?
would google news be blank after a while?
just wanted to point out: the internet in asia broke due to the earthquake in taiwan.