Stuck in a traffic jam? This is how your Internet data avoids gridlocks

Can your internet data get stuck in a virtual traffic jam?

So much of modern life depends on fast transportation. In ancient Rome, traveling between cities used to take days or weeks, while nowadays people commute regularly for hundreds of kilometres every day to go to study or work. An even more extreme jump occurred in the speed of communication. While delivering a letter used to take at least as long as human travel, requiring physical transport and delivery, we can now connect instantly with people on the other side of the world and exchange a huge amount of data, including photos, music, videos and blog posts like this one.

Everybody that has ever commuted in the modern world knows that, as more and more people travel, traffic jams become a common and frustrating experience. What you may not know is that, in the early days of the Internet, it used to be common for data to get stuck in a virtual equivalent of a road gridlock. In the late ‘80s, the situation became so bad that some universities saw their Internet connection speed reduced by a factor of a thousand times, to the point that it would take one second to send or receive five English characters. Receiving a full-size tweet would have required you a full minute of waiting, while a website with a single medium sized picture would have taken around one hour to be loaded.

The increasing occurrence of these data jam events inspired researchers Mike Karels and Van Jacobson from the University of California to search for solutions to the problem. They noticed that, as each Internet client was trying to push as much data as possible into the network, data packets would start to get lost. When data is lost, it needs to be sent again, leading to waste of network bandwidth and a significant drop in network speed. To avoid this situation, they came up with a strategy that did not require to modify the expensive physical network infrastructure, but only a small change of behaviour on the client side. Instead of sending as much data as possible, a client should carefully probe the network capacity, gradually increasing its sending speed until some data begins to be lost. At this point, instead of immediately trying again, the client should cut their speed in half, allowing for the network to not be overused. The client will then increase their speed again, until the next data loss, repeating the cycle over and over. The researchers demonstrated that this strategy allowed for the network to be used at close to full capacity without creating data jams that would significantly reduce network speed.

This research was revolutionary at the time, allowing the Internet to grow exponentially and connect everybody on the planet like never before. But one question remains open. When can science find a solution to the real life commute traffic jams everybody hates? Progress is always ongoing, and traffic aware self-driving cars may become part of the revolution. But this is a topic for another day.

Manuel Furia

Source:

V. Jacobson. 1988. Congestion avoidance and control. In Symposium proceedings on Communications architectures and protocols (SIGCOMM ’88). Association for Computing Machinery, New York, NY, USA, 314–329. https://doi.org/10.1145/52324.52356