Akamai: From Theory to Practice
The development of Akamai Technologies is a classic MIT story. Motivated by curiosity, Professor Tom Leighton’s search for fundamental knowledge led to the creation of a company that would soon revolutionize the Internet. As one of the world’s preeminent authorities on algorithms for network applications, Leighton’s work behind establishing Akamai was based on recognizing that a solution to freeing up web congestion could be found in applied mathematics and algorithms. The research that spawned Akamai began in MIT’s Mathematics Department and Laboratory for Computer Science in 1995. Today, Akamai’s global platform of thousands of specially-equipped servers helps the Internet withstand the crush of daily requests for rich, dynamic content. On October 30, 2013, Leighton talked about the founding of the company and the challenge of being the leading cloud platform for helping enterprises provide secure, high-performing user experiences on any device, anywhere. The Dean’s Colloquium with Tom Leighton was presented by the School of Science.
At MIT, when a scientist discovers something exciting, it doesn't just languish in the laboratory... it actually goes out and changes the world. -Marc Kastner, former Dean
Leighton began the research that laid the foundation for Akamai in 1995, working on the mathematics of routing through various nodes to endpoints. At the time, he wasn’t thinking of how his work might be applied to the real world, much less the Internet. His goal was the pursuit of knowledge: as Leighton told the audience, “I did not write any code; I wrote papers.” The Internet itself hardly qualified as a real world application in the mid-nineties anyway. People were still trying to imagine how to integrate the Internet into the business of everyday life and what infrastructure would be needed to support it. Meanwhile, the Internet could only be a novelty with more promise than practical uses – even when it managed to work. Yet Leighton needed funding, and while working as the “token mathematician” at DARPA, he was advised that one way to attract grant money was apply his algorithms to the logistics of the Internet.
Luckily, Tim Berners-Lee at the Algorithms Group at the MIT Laboratory for Computer Science had a theory that “hot-spots” could cause problems for the Internet, crashing sites when there was a very high-traffic event. Very few people thought that hot-spots were a serious problem at the time, since there simply wasn’t enough traffic at any one time to challenge the existing infrastructure. Leighton’s grad student, Danny Lewin, was already working on a way to support load balance called “hashing,” a way to store objects that is fast, but is vulnerable to breaking when the number of servers changes or servers go down. Together, Leighton and Lewin wrote a paper that laid out a solution to the problem of hot-spots and proposed it to an applied mathematics conference. The paper’s potential was overlooked, and it was rejected as too theoretical. Leighton and Lewin’s idea would fare better at the MIT 50K Competition in 1997, where, although it didn’t win, they attracted the notice of some venture capitalists and began to receive valuable feedback from major websites as they developed their prototypes and their business plan.
But what Leighton and Lewin needed was proof that hot-spots were going to be a significant obstacle to a useful and reliable Internet – and a chance to show the world that they had a good solution for them. As chance would have it, they would get exactly what they needed. Their next stroke of luck came in the form of a Superbowl advertisement for the first live Internet broadcast of the Victoria’s Secret runway show in 1999.
The advertisement generated so much buzz that people were flocking to the Victoria’s Secret website to see the show. The traffic took down Internet service in most of Texas and nobody anywhere in the world got to see the show. Everyone knew then that hot-spots were actually a problem. Leighton and Lewin soon got their first piece of content to test their algorithms on: a single pixel buried deep within a Disney site. Their next two big breaks came that same year. Concerned by the Victoria’s Secret site crash, InfoSeek was looking for a way to make sure they could keep the ESPN site they were hosting from crashing during March Madness. Akamai agreed to test its product on Infoseek at up to 2000 hits per second, even though no one at Akamai really knew for sure that they could handle that volume. Up until that point it had only been tested at one hit every couple of minutes (that is, whenever someone at Akamai got around to physically hitting a key). March Madness drove InfoSeek’s traffic up to 3000 hits per second and their site did in fact crash. But thanks to Akamai, nobody ever knew about the crash. In fact, access speeds had been increased five-fold.
The next break came about when Steve Jobs was looking to promote QuickTime by obtaining exclusive rights to the Star Wars trailer. Apple’s site created a hot-spot and crashed, but that didn’t mean no one got to see the trailer. Entertainment Tonight had contacted Akamai to support their site and released their own bootleg copies of the trailer the same day. Again thanks to Akamai, Entertainment Tonight avoided a crash. Jobs was impressed and became Akamai’s first strategic investor. Apple still relies on Akamai to support such features as FaceTime.
By 2010, Akamai exceeded $1 billion in revenue, although the road to its success was far from smooth. By 2001, Akamai was worth $35 million on paper, largely due to the stock bubble, but it was bleeding money and paying its employees, mostly MIT students, in stock options. When the bubble burst, Akamai had to bail its employees out, who were rich on paper, poor on cash, and burdened by impossibly large tax obligations. Akamai’s debt became junk debt. Moreover, in 2001 Akamai lost one of its founders, Danny Lewin, a former captain of an elite Israeli force, who happened to be on one of the hi-jacked planes on 9/11 and died attacking one of the hi-jackers. No one thought Akamai would survive. And yet, it did. Leighton attributes Akamai’s survival through the lean years of the early 2000’s and its present success to MIT’s unique culture, where people work hard and work together to solve problems. Akamai is now looking forward to reaching $5 billion in revenue in the next few years through meeting some of the emerging obstacles in bringing high-quality content to consumers, quickly, seamlessly, and securely.