In my worldbuilding, we're looking at the typical dystopian setup of most of humanity being in some way dead and gone. I'm not looking for what would happen if the internet shut down in general—that seems to have been answered more than enough. But what happens to the data on the internet? If, say, people managed to band together and find a way to restore internet access for the remnants of humanity (assuming power/equipment), would all of the information currently stored on the internet be gone due to the systems having been shut down? Would it ever be possible to recover any of the old information from the internet, assuming you could access the internet (as a whole or in part) again?
-
$\begingroup$ there are redundant repositories of data serving archive.org $\endgroup$– AllanCommented Dec 17, 2021 at 17:42
-
$\begingroup$ BTW, welcome to Worldbuilding! Please be sure to take the tour and read through the FAQ -- not only will you better understand how Stack Exchange sites (like this one) work, you'll get some free reputation out of it. $\endgroup$– Zeiss IkonCommented Dec 17, 2021 at 17:45
-
2$\begingroup$ It would've probably helped a lot/answered your question to read a brief summary of what the internet actually is before asking this. $\endgroup$– NotThatGuyCommented Dec 18, 2021 at 2:13
-
2$\begingroup$ It might be helpful to take a step back and actually study how computers and networks work. It seems like you're coming at this with a pretty big technological blind spot and with a lot of misconceptions about what the internet actually is (and what it isn't). $\endgroup$– J...Commented Dec 18, 2021 at 15:44
-
3$\begingroup$ "Restoring the data" isn't going to be the problem. Restoring power and connectivity, as well as booting and admin issues to the millions of servers in hundreds of thousands of different locations is going to be the show-stopper. It will take an unimaginable number of person-hours of (skilled) labor to do this. $\endgroup$– RBarryYoungCommented Dec 18, 2021 at 17:45
9 Answers
The "Internet" refers mainly to the network connection that makes millions of computers accessible from any connected device. The actual data, however, is stored on millions of mass storage devices, ranging from small platter or SSD volumes to optical or magneto-optical carousels containing petabytes in a single location.
None of that data would be significantly compromised by loss of the network backbones; all that would be required to make it accessible again is to restore high-rate data connections between a substantial fraction of the machines that mount and read/write those storage volumes.
-
3$\begingroup$ don't forget to restore power to those machines that mount and read/write those storage volumes. $\endgroup$ Commented Dec 17, 2021 at 22:24
-
4$\begingroup$ Considering this would include physically going to these locations and getting (at least the local) power grid to work again, I wouldn’t consider this trivial $\endgroup$ Commented Dec 18, 2021 at 2:17
-
14$\begingroup$ A "black start" of the Internet would be a fair bit more difficult than just turning everything back on. There are quite a few systems that depend on other systems, and I wouldn't be surprised to find some dependency loops out there. $\endgroup$– MarkCommented Dec 18, 2021 at 3:43
-
2$\begingroup$ One should also consider the time aspect. If the internet is down for a few hours this is probably correct. If the internet and the accompanying infrastructure is down for years I don't think one could just restart. $\endgroup$– quaragueCommented Dec 18, 2021 at 12:13
Don't forget you'll need to deal with passwords. Even the if you power on a server and all the bytes in the disks survived you'll need to deal with decrypting any encyrpted data or having login passwords. It's not as if you'll be able to a password reset as no one's email will be working.
-
3$\begingroup$ Of the current answers, this one is closest to the reality for warehouse-sized computers or datacenters. All of the data is durably backed up to long-term tape drives or WORM media, but encrypted. Restarting a datacenter involves restarting the control room, which might boil down to a single SD card full of cryptographic key material locked by a passphrase. $\endgroup$– CorbinCommented Dec 18, 2021 at 3:47
-
$\begingroup$ @Corbin Don't forget the physical access controls (PINs, biometrics, etc.) needed to get into the datacenters in the first place. $\endgroup$ Commented Dec 18, 2021 at 21:36
-
3$\begingroup$ Don't worry, the data center passwords are on a post-it note in the top left drawer of the desk of the assistant to the manager. Or at least they were there yesterday... $\endgroup$ Commented Dec 19, 2021 at 1:36
-
$\begingroup$ That may depend on the story, to en extent. No widely used encryption is unbreakable, only unfeasible to decipher, but if we assume e.g. aliens with advanced technology, they may be able to break the passwords and eventually read the data. The issue of the black start stays, though. $\endgroup$– FraxCommented Dec 19, 2021 at 20:25
-
$\begingroup$ @AlexHajnal I have only been in a few datacenters, but I personally do not believe that the typical datacenter's physical security is a serious obstacle. A writer might choose to give their datacenter armed guards/traps and reinforced doors, of course. $\endgroup$– CorbinCommented Dec 19, 2021 at 23:19
"There is no cloud, there are only other people's computers"
As pointed out by Zeiss Ikon, data isn't on "the internet," it is on computers. But I expect significant problems, up to and including data loss, when communications go down and later get restored. Programmers are making mistakes all the time, not thinking of possible errors, and so programs are buggy. The recent Log4Shell exploit is a good example where a possibly beneficial function could be abused in unexpected ways.
- I would expect that a number of caches have to be rebuilt, with a loss of some recent data.
- Badly made systems could be impossible to restart, with component A relying on component B, component B relying on component C, and the most recent version of component C relying on component A (even if the original C didb't).
It depends on how long does it pass between the shutdown and the restart: the data that we see in the cloud is stored in some data centers around the globe, and these data centers, if not physically destroyed, will keep the stored data as long as the memory allows.
Solid state memory stores data under the form of localized charges, and when the amount of this charge is changed, the data is corrupted. This can happen either because of electrostatic discharge (which also physically damages the components) or by charges present in the environment, typically supplied by cosmic rays and radioactive decay.
The more the data storage is exposed to these sources, the higher the chances of data corruption.
Data loss will be unavoidable
If internet is abandoned for a relevant amount of time, with no way to guard loss of information, because there has been no access and no maintenance personnel, you'll have issues. Things will be lost forever. This is already what happens now.. information not accessed will get lost. Without active internet, your DNS servers will get no requests. At some point, DNS servers will also cease to communicate URLs between them, to keep up the DNS database. When too many DNS servers get lost, the backbone of the information (URL's and links) will disappear.
As L.Dutch made clear, it would depend on the amount of time past, what would happen with SSD like storage. Any information on magnetic media will get wiped out in 30-50 years, whether machines are switched off or not. A spinning harddisk wears out even faster. Normally, long before a harddisk has issues, the datacenter will move data, replace it, and move data back.. as long as there is personnel. In your scenario, data will really get lost.
I can add the importance of the close down scenario.. how did the internet shut off ? how well were files preserved during that event ?
-
1$\begingroup$ "At some point, DNS servers will also cease to communicate URLs between them, to keep up the DNS database." – DNS lookups and zone transfers should work fine indefinitely as long as there's network connectivity when a query is made (provided all server addresses remain the same). The only exception to this is if the zone is using DNSSEC; in that case then lookups for the zone will fail when its DNS server's certificate expires. And a nitpick: DNS works with domain names (e.g.
worldbuilding.stackexchage.com.
), not URLs (e.g.https://worldbuilding.stackexchange.com/questions/220413/
). $\endgroup$ Commented Dec 18, 2021 at 22:21 -
$\begingroup$ @AlexHajnal the expiration is the issue.. when there is 1) no transfer and 2) the registration is expired, DNS will unregister the link automatically. After 5-10 years you'll end up without any registered domain. Of course, when the internet is reactivated, DNS will work again, but I think all domains will have to be reactivated first. Manually, that is register and pay for the domain. All content without active owner and webmin will have vanished, or.. it can't be found at the original IP. The shutdown could have lasted for decades..., this is part of a recovery from a dystopian episode. $\endgroup$– GoodiesCommented Dec 18, 2021 at 22:35
-
$\begingroup$ Good point, I hadn't considered the registrars as a point of failure. You'd be left with a rather restricted set of resolvable domains in that case. I'd expect most government (≥ state/province level) domains wouldn't expire but anything on
*.org.
,*.co.uk.
, etc. would become unresolvable. One other consideration is that domains can only be registered for up to 10 years at a time; longer terms sold by registrars (e.g. GoDaddy forcom.
) involve periodic re-registration with the TLD registry operator (e.g. VeriSign forcom.
, this is handled by the domain's registrar e.g. GoDaddy). $\endgroup$ Commented Dec 18, 2021 at 23:21 -
$\begingroup$ When the internet is rebooted after say 30 years, lots of new content will arrive quickly on a few URLs, the old content will take time. You'd get the domain name wars all over again ! At first, any ownership is free to grab, that will happen in the first few weeks. Popular items like Wikipedia can be restored, because users keep database backups everywhere. But social media, forums, blogs, science, commercial and newspaper websites will have issues with old content, because many surviving stuff consists of cache data that will have to be put online as is, or reverse engineered. Manually. $\endgroup$– GoodiesCommented Dec 18, 2021 at 23:24
The data is stored on devices that, 99.9% of the time, withstand a power failure, and work just fine once power is restored.
However, leaving such devices alone for a long time can make them unreadable. I have tried to access a hard drive I had stored on a shelf for two years, I was just going to see if I could use it for something else. Despite being undisturbed for those years, it was no good anymore, the only thing I could use for was a paperweight or doorstop.
There is also the problem of the network itself. It takes software to access web pages, unless you are going to engage in a lot of searching for javascripts and templates and other stuff, the raw data of web pages is not easy to figure out. An internet page may read both code and data from remote devices; when you read Amazon, it is building the page code it sends you on the fly from dozen of sources. If any of these sources are missing the page may not render at all.
This interdependency across the network can mean the whole Internet degrades rather quickly, because many elements can be affected by each hardware failure as time goes on.
If it goes down for a few weeks, I'm sure it will come back fine. If it goes down for a decade, I doubt it would come back fine.
-
1$\begingroup$ This. I was testing some drives that had been sitting around for some years. Half won't work. $\endgroup$ Commented Dec 20, 2021 at 16:02
The biggest concern IMO is whether the Domain Name System (DNS) remains functional. DNS is responsible for converting names (e.g. worldbuilding.stackexchange.com
) to addresses (e.g. 151.101.193.69
). I see a few potential problems here:
DNS is hierarchical so if the top-level servers (e.g. those for
com.
) are unreachable or down then no addresses can be looked up.The DNS servers for each domain (and subdomain) are maintained by the site owners. If the server addresses change when the Internet is brought back up and the owners aren't there to update the DNS records then the sites at those domains will be unreachable even if the servers are up.
As pointed out by @Goodies in the comments, domain registrations expire and when they do the registrars1 typically deregister them (or take ownership themselves). Even if automatic deregistration doesn't happen, domain names can only be registered for 10 years at a time2 after which the domain must be re-registed with the TLD's registry operator3. If this doesn't happen then the domain will likewise be deregistered4. In this case I'd expect only government domains (at or above the state/province level) to continue to be resolvable over the long term with most domains (
*.org.
,*.co.uk.
, etc.) becoming unresolvable.Related: The Internet is not a single place but rather many places (servers) all over the world. It's not enough to just restore access to your ISP, you also need to restore (at least some of) your ISP's links to other providers, those providers' peering links, etc.
1 Registrars (such as GoDaddy or TuCows for com.
) sell domain names at retail/wholesale.
2 At least for some Top Level Domains (TLDs), possibly for all.
3 I'm not sure whether this is true for all registry operators.
A registry operator runs the actual domain name servers for a TLD (e.g. VeriSign for com.
). Registrars notify the operator when changes are made to individual domains' NS
(nameserver) records.
4 Again, I'm not positive that this is the case for all registry operators.
Probably not all of it.
Some systems are "trivial". Think of your traditional servers and networking hardware. Just power on all the equipment and they have running configuration saved and automatically restart needed things. Make sure that interconnections are also powered on that is every piece of equipment on the way between them. This is however forgetting durability and other such concerns.
On other hand specially with cloud and very large systems this is unlikely to be enough. I think getting up your cloud platform is rather involved process from cold-start requiring very specialised and probably even undocumented knowledge. Needing things done in just right order with probably some caressing and giving time to get certain programs running and stable before starting next ones.
And at that point you might not be even running the stuff we think of as Internet there. Those thousands or hundreds of thousand individual web sites, servers and others might need similar actions to get up and running. Or in worst case they might not even be recoverable. Our modern software development some times isn't exactly stable or great. And likely mean that you need some other system up and running or even some random laptop somewhere from where you run other software to upload the software and run it in cloud.
All the steps above might be needed to access some of the data. Or they might not even be enough.
In addition to the catastrophic DNS problems Alex Hajnal mentions there is a related problem with certificates. Within a few years everybody's certificate is expired, nothing that uses https works.
In theory this might be averted by setting the clocks wrong--but the people restoring it likely wouldn't know this and very well might not even know what time to set it to.
And even if they get it back up, those certificates will be ticking time bombs--the people with the passwords to update them are gone.