Perenet based on pigeons is still the fastest way to transfer large amounts of information.

A carrier pigeon loaded with microSD cards is capable of transferring large amounts of data faster and cheaper than almost any other method.

Note perev .: even though the original of this article appeared on the IEEE Spectrum website on April 1, all the facts listed in it are quite reliable.

In February, SanDisk announced the release of the world's first terabyte microSD flash card. It, like other cards of this format, is tiny, measuring just 15 x 11 x 1 mm, and weighs 250 mg. It can fit an incredible amount of data in a very small physical space, and you can buy it for $ 550. So that you understand, the first 512 GB microSD cards appeared just a year before, in February 2018.

We are so accustomed to the speed of progress in the field of computer technology that these increases in drive density remain virtually unattended, and sometimes get a press release and a couple of blog articles. More interesting (and likely to lead to more serious consequences) is how much faster our ability to generate and store data is growing compared to our ability to transfer it over networks accessible to most people.

This problem is not new, and for several decades various types of hitronets have been used to physically transport data from one place to another - on foot, by mail, or by more exotic methods. One of the methods of data transfer, which has been actively used for the past thousand years, is carrier pigeons, capable of traveling hundreds or even thousands of kilometers long, returning home, and using navigation techniques, the nature of which has not yet been precisely studied. It turns out that in terms of bandwidth (the amount of data transmitted over a given distance for a certain time), “peronet” based on pigeons remains more efficient than typical networks.


From the "standard for the transmission of IP datagrams by air carriers"

on April 1, 1990, David Weitzman proposedThe Internet Engineering Council Request for Comment (RFC), called the Air Transport IP Datagram Standard , now known as IPoAC. RFC 1149 describes the “experimental method of encapsulating IP datagrams in air carriers,” and it has several updates regarding both the quality of service and the transition to IPv6 (published April 1, 1999 and April 1, 2011, respectively).

Sending RFC on April Fool's Day is a tradition that began in 1978 with RFC 748, which suggested that after sending the IAC DONT RANDOMLY-LOSE command to the telnet server, the server would stop randomly losing data. Pretty good idea, right? And this is one of the properties of April Fool’s RFC, explains Brian Carpenter, who led the CERN Networking Working Group from 1985 to 1996, chaired the IETF from 2005 to 2007, and now lives in New Zealand. “It must be technically feasible (that is, not violate the laws of physics), and you must read at least a page before you realize that this is a joke,” he says. “And, of course, he must be absurd.”

Carpenter, along with his colleague Bob Hinden, themselves wrote April Fools' Day RFCs, which described the IPoAC upgrade for IPv6 in 2011. And even two decades after its introduction, IPoAC is still well known. “Everyone knows about air carriers,” Carpenter told us. “One day, Bob and I talked at an IETF meeting about the spread of IPv6, and the idea of ​​adding it to IPoAC came up very naturally.”

RFC 1149 , which originally defined IPoAC, describes the many benefits of the new standard:

Many different services can be provided by prioritizing pecking. Additionally, there is a built-in recognition and destruction of worms. Since IP does not guarantee 100% packet delivery, the loss of carrier can be reconciled. Over time, carriers recover themselves. Broadcast is not defined, and a storm can lead to data loss. It is possible to make persistent delivery attempts before the carrier falls. Audit traces are generated automatically; they can often be found in cable trays and on logs . log means both “log” and “log for records” / approx. perev. ].

The Quality Improvement Update (RFC 2549) adds several important details:

Multicasting, although supported, requires the implementation of a device for cloning. Carriers may get lost if they are located on a tree being cut down. Carriers are distributed across the inheritance tree. TTL carriers averaged 15 years on average, so their use in expanding ring searches is limited.

Ostriches can be considered as alternative carriers, with much greater ability to transfer large amounts of information, but providing slower delivery and requiring bridges between different areas.

An additional discussion of the quality of service is available in the Michelin Guide .

An update from Carpenter describing IPv6 for IPoAC, among other things, mentions potential difficulties with packet routing:

The passage of carriers through the territory of carriers similar to them, without establishing agreements on equitable information exchange, can lead to a sharp change of route, looping packages and delivery out of order. The passage of carriers through the territory of predators can lead to a significant loss of packages. It is recommended that these factors be considered in the routing table compilation algorithm. Those who will implement these routes in order to guarantee reliable delivery should consider routing based on policies that bypass areas with a predominance of local and predatory carriers.

There is evidence that some carriers tend to eat other carriers and further transport the eaten payload. Perhaps this will serve as a new method for tunneling IPv4 packets into IPv6 packets, or vice versa.

The IPoAC standard was proposed in 1990, but messages with carrier pigeons were sent much longer: the photo shows the sending of a carrier pigeon in Switzerland, between 1914 and 1918.

It is logical to expect from the standard, the concept of which was invented back in 1990, that the original format for data transmission IPoAC protocol was associated with printing hexadecimal characters on paper. Since then, much has changed, and the amount of data that fits into the given physical volume and weight has increased incredibly, despite the fact that the payload of an individual pigeon remains the same. Pigeons are capable of carrying a payload of a significant percentage of their body weight - the average carrier pigeon weighs about 500 grams, and at the beginning of the 20th century they could carry 75-gram cameras for reconnaissance on enemy territory.

We talked with Drew Lesofsky , a lover of pigeon races from Maryland, and he confirmed that pigeons can easily carry up to 75 grams (and maybe a little more) "during the day at any distance." At the same time, they can fly a considerable distance - one fearless bird holds the world record for the carrier pigeon, who managed to fly from Arras in France to her home in Ho Chi Minh City in Vietnam, having covered 11,500 km in 24 days. Most carrier pigeons, of course, are not able to fly that far. A typical length of a long race track, according to Lesofsky, is about 1000 km, and birds overcome it at an average speed of about 70 km / h. At shorter distances, sprinters can reach speeds of up to 177 km / h.

Putting all this together, it can be calculated that if we load the carrier pigeon up to its maximum load capacity of 75 grams with 1 TB microSD cards, each of which weighs 250 mg, then the pigeon will be able to carry 300 TB of data. Having crossed the path from San Francisco to New York (4,130 km) at maximum sprint speed, he would have reached a data transfer rate of 12 Tb / h, or 28 Gb / s, which is several orders of magnitude higher than most Internet connections. In the USA, for example, the fastest average download speed is observed in Kansas City, where data is transmitted via Google Fiber at a speed of 127 Mbps. At such a speed, it would take 240 days to download 300 TB - and during this time, our pigeon would be able to fly around the globe 25 times.



Suppose this example does not look very realistic, because it describes some kind of super-blue, so let's slow down. Take an average flight speed of 70 km / h, and load the bird at half the maximum load in terabyte memory cards - at 37.5 grams. And still, even if we compare this method with a very fast gigabit connection, the pigeon wins. A pigeon will be able to circumnavigate more than half of the globe in the time until our file transfer ends, which means that it will be faster to send data to a pigeon literally anywhere in the world than to use the Internet to transfer them.

Naturally, this is a comparison of pure bandwidth. We do not take into account the time and effort to copy data to microSD cards, load it onto a pigeon, and read data on the arrival of a bird at its destination. The delays are obviously high, so anything other than one-way transmission will be impractical. The biggest limitation is that the carrier pigeon flies in only one direction and one destination, so you cannot choose the purpose of sending the data, and you also have to transport the pigeons to where you are going to send them, which also limits their practical use .

However, the fact remains - even with realistic estimates of the payload and speed of the pigeon, as well as the Internet connection, the net throughput of the pigeon is not easy to surpass.

Given all this, it is worth mentioning that the transmission of data by pigeons was checked in the real world, and they coped well with this. A group of Bergen Linux users from Norway in 2001 successfully implemented IPoAC , sending one ping with each pigeon to a distance of 5 km:

Ping was sent at about 12:15. We decided to make an interval of 7.5 minutes between packets, which ideally should lead to a couple of packets remaining unanswered. However, everything went wrong. At our neighbor a flock of pigeons flew over the site. And our pigeons did not want to fly straight home, at first they wanted to fly with other pigeons. And who can blame them for this, given that the sun came out for the first time after a couple of cloudy days?

However, their instincts won, and we saw how, having frolic for about an hour, a couple of pigeons broke away from the pack and headed in the right direction. We rejoiced. And it really was our pigeons, because soon after that we received a report from another point that the pigeon landed on the roof.

Finally, the first pigeon arrived. The data packet was carefully removed from its paw, unpacked and scanned. After manually checking the OCR and fixing a couple of errors, the package was accepted as valid, and our glee continued.

For really large amounts of data (such that the required number of pigeons will become difficult to maintain), physical methods of moving still have to be used. Amazon offers Snowmobile , a 45-foot truck transport container. A single Snowmobile can carry up to 100 PB (100,000 TB) of data. It will not move as fast as the equivalent flock of several hundred pigeons, but it will be easier to work with.

Most of the people, apparently, are satisfied with an extremely leisurely download, and they are not very interested in investing in their own carrier pigeons. This really requires a lot of work, says Drew Lesofsky, and the pigeons themselves usually behave, not like data packets:

GPS technology is increasingly helping pigeon racing enthusiasts, and we get a better idea of ​​how our pigeons fly and why some fly faster than others. The shortest line between two points will be a straight line, but pigeons rarely fly in a straight line. They often draw zigzags, flying in approximately the right direction, and then adjusting course, approaching the destination. Some of them are physically stronger and fly faster, but a pigeon that is better oriented, has no health problems and is physically trained, can outrun a fast flying pigeon with a bad compass.

Lesofsky trusts pigeons enough as data carriers: “I would quite confidently send information with my pigeons,” he says, while taking care of error correction. “I would issue at least three at once to ensure that even if one of them has a bad compass, the other two will have better compass, and in the end the speed of all three will be higher.”

Problems with the implementation of IPoAC and the increase in the reliability of fast enough (and often wireless) networks mean that most of the services that relied on pigeons (and there were many) have switched over to the more traditional methods of data transfer over the past few decades.

And because of all the preliminary preparations necessary for equipping the data transmission system with pigeons, comparable alternatives (like fixed-wing drones) can become more viable. However, pigeons still have some advantages: they scale well, work for seeds, are more reliable, they have a very sophisticated system for avoiding obstacles both at the software level and at the iron level, and they are able to recharge themselves.

How will all this affect the future of IPoAC? There is a standard, it is available to everyone, albeit a little absurd. We asked Brian Carpenter if he was preparing next updates to the standard, and he said that he was thinking about whether the pigeons would be able to transfer qubits. But even if IPoAC is a bit complicated (and a little silly) for your data transfer needs, all sorts of non-standard communication networks will remain necessary for the foreseeable future, and our ability to generate huge amounts of data continues to grow faster than our ability to transfer them.

Thanks to the user AyrA_ch for pointing out the information with his post on Reddit , and for the convenient IPoAC calculator , which helps to calculate how truly pigeons are ahead of other data transfer methods.