The internet is broken. You can blame sharks. And Netflix
Bad news: The internet ran out of space on Tuesday. Worse news: Sharks are eating what is left.
Thankfully, it is not as bad as it sounds. Yet. But there are some existential threats to the internet on the horizon, and there are worse ways of putting it than to point out that the whole thing is full up.
On Tuesday, a small collection of out-of-date routers, in charge of mapping out routes using the internet’s backbone of physical cables, failed. The routers use a system called the border gateway protocol, or BGP, to track these routes, of which there are around half a million.
To be specific, there are actually 512,000, the magic number that caused the failures earlier this week. It turns out that is the absolute maximum number of routes some older hardware can hold. Some routers slow to a crawl, while others simply forget the extra routes.
That was the cause of the outages, which managed to affect sites as large as eBay. US internet service provider Verizon added 15,000 more routes to the BGP, and the dependable routers finally fell over, five years after experts say they should have been retired from service.
To add insult to injury, on Wednesday, reports surfaced that Google was having to protect what was left of the internet’s backbone from shark attacks. Unlike so many things named by the architects of the internet, “shark attack” is not just a catchy name for something dull: it is literally sharks attacking the undersea cables that connect the world.
Sharks are, literally, attacking the internet. Photograph: Alamy
Apparently sharks are attracted to the magnetic field created by the high voltage carried through newer undersea cables and, thinking they’re fish, they bite them. Then, if underwater footage is anything to go by, they realise their mistake and swim away from the damage they have caused.
There is some good news. The cables are now largely shark-proof, and the old routers that failed on Tuesday will now be out of service, unable to connect to the new network. Those headlines about the internet being “full up”? Unfortunately, they weren’t quite as ridiculous as they sounded.
For one thing, our ability to gobble up bandwidth continues to grow as fast as our ability to install new infrastructure: people using Netflix and YouTube accounts for more than half of US internet traffic. As the average American watches five hours of broadcast television a day, this percentage is only going to grow, as more content is delivered via the internet.
But, as problems go, the solution to that one is fairly simple: throw money at it, and build more and better cables. By another measure, though, the internet is full in a way that is much more difficult to fix.
We are about to run out of IP addresses, the short 12-digit identifiers for every single device online. A cap of 4.2bn looked fine back in the 1970s, when the protocol was being put together; it looked a bit shaky in the 80s, as the personal computer was born; and was obviously doomed by the 90s, once mass adoption of the internet became a reality. The noughties brought smartphones, using even more addresses, and that was before the much heralded “internet of things”.
A fix is on the horizon: IPv6, which ups the limit to something in the order of 300 trillion trillion trillion devices, which should hopefully be enough for quite some time. But IPv6 requires every single device on the connection to switch over – and it is not backwards-compatible. That means that it is difficult to upgrade in a piecemeal fashion: it is all or nothing, and once a device switches over to IPv6, it can’t talk to things using the old connection.
That explains why there has been such a delay in flicking the switch, even as the internet steadily fills up. In 2011, APNIC, which is in charge of allocating IP addresses in the Asia Pacific region, ran out of new addresses, and the other registries will soon follow.
Well, at least we’ve solved the shark attack.
Thankfully, it is not as bad as it sounds. Yet. But there are some existential threats to the internet on the horizon, and there are worse ways of putting it than to point out that the whole thing is full up.
On Tuesday, a small collection of out-of-date routers, in charge of mapping out routes using the internet’s backbone of physical cables, failed. The routers use a system called the border gateway protocol, or BGP, to track these routes, of which there are around half a million.
To be specific, there are actually 512,000, the magic number that caused the failures earlier this week. It turns out that is the absolute maximum number of routes some older hardware can hold. Some routers slow to a crawl, while others simply forget the extra routes.
That was the cause of the outages, which managed to affect sites as large as eBay. US internet service provider Verizon added 15,000 more routes to the BGP, and the dependable routers finally fell over, five years after experts say they should have been retired from service.
To add insult to injury, on Wednesday, reports surfaced that Google was having to protect what was left of the internet’s backbone from shark attacks. Unlike so many things named by the architects of the internet, “shark attack” is not just a catchy name for something dull: it is literally sharks attacking the undersea cables that connect the world.
Apparently sharks are attracted to the magnetic field created by the high voltage carried through newer undersea cables and, thinking they’re fish, they bite them. Then, if underwater footage is anything to go by, they realise their mistake and swim away from the damage they have caused.
There is some good news. The cables are now largely shark-proof, and the old routers that failed on Tuesday will now be out of service, unable to connect to the new network. Those headlines about the internet being “full up”? Unfortunately, they weren’t quite as ridiculous as they sounded.
For one thing, our ability to gobble up bandwidth continues to grow as fast as our ability to install new infrastructure: people using Netflix and YouTube accounts for more than half of US internet traffic. As the average American watches five hours of broadcast television a day, this percentage is only going to grow, as more content is delivered via the internet.
But, as problems go, the solution to that one is fairly simple: throw money at it, and build more and better cables. By another measure, though, the internet is full in a way that is much more difficult to fix.
We are about to run out of IP addresses, the short 12-digit identifiers for every single device online. A cap of 4.2bn looked fine back in the 1970s, when the protocol was being put together; it looked a bit shaky in the 80s, as the personal computer was born; and was obviously doomed by the 90s, once mass adoption of the internet became a reality. The noughties brought smartphones, using even more addresses, and that was before the much heralded “internet of things”.
A fix is on the horizon: IPv6, which ups the limit to something in the order of 300 trillion trillion trillion devices, which should hopefully be enough for quite some time. But IPv6 requires every single device on the connection to switch over – and it is not backwards-compatible. That means that it is difficult to upgrade in a piecemeal fashion: it is all or nothing, and once a device switches over to IPv6, it can’t talk to things using the old connection.
That explains why there has been such a delay in flicking the switch, even as the internet steadily fills up. In 2011, APNIC, which is in charge of allocating IP addresses in the Asia Pacific region, ran out of new addresses, and the other registries will soon follow.
Well, at least we’ve solved the shark attack.
0 comments:
Post a Comment