An unassailable network 17 years in the making.
Coming down Broadway Street from inner Cambridge, Akamai is the gatekeeper to Boston’s oldest and most mature innovation district.
Akamai is more than just the gatekeeper to Kendall Square. The world’s preeminent content delivery network is the backbone at the heart of the Internet itself. Akamai is a product of the Web and intrinsically ingrained into it. Nobody really thinks of it anymore, but Akamai is one of the Boston areas’ most preeminent startups of the Web era.
Of course, Akamai is no longer a startup. By Web standards, Akamai is an old, scarred veteran that lived through the Dot-Com bust and the Great Recession, hanging on by the strength of its product and astute management to become one of the fundamental backbones of the Internet and the traffic that flows through its pipes.
On any given day, between 15% and 30% of the traffic on the Internet flows through Akamai’s nodes. Akamai knows more about how the world connects to the Internet on a second-to-second basis than probably any other company in existence. To move content around the Internet, Akamai is on 1,400 networks in 100 countries and 700 cities around the world, running through 200,000 servers. The system took 17 years to create and is virtually unassailable in the market.
“I think the architecture, think of our servers, in as many locations as possible. We are in thousands of locations and networks,” said Akamai cofounder and CEO Dr. Tom Leighton in an interview with ARC. “And our software is on over 100 million end user devices where they’ve given us the permission to use the device to help them deliver traffic. We want it to be ubiquitous. The more distributed it can be, the more efficient it gets, the better performance, lower cost, better scale.”
Akamai’s algorithmic, software-defined network knows no single point of failure or even multiple points of failure. It is designed to be intelligent and robust, redundant and secure.
But like so many other companies that were born and matured in the late 1990s, Akamai almost folded when the Web bubble burst.
Theoretical Mathematics At MIT Leads To Akamai
The story of Akamai must be told on a few parallel tracks: the personal, the technological and the financial.
Akamai was born out of the Massachusetts Institute of Technology as a project between a student and a professor. Danny Lewin—the student—worked with his professor—Dr. Tom Leighton—to create algorithms that would speed up the delivery of content over the Internet. Lewin, Leighton and another grad student named Preetish Nijhawan took the concept of Akamai to MIT’s $50k Entrepreneurship Competition and won in 1997. In 1998 Lewin would win the Morris Joseph Lewin Award for Best Masterworks Thesis.
“I think the mathematics and the mathematical background is what has allowed us to be successful,” Dr. Leighton said. “It is a necessary component. I think the culture is also a necessary component. You know, tremendous drive, a sense of urgency, transparency. A culture like what you find at MIT. The belief that you can solve pretty much any problem by really working hard at it, working as a team.”
The first big wins for Akamai came in 1999. The trailer for the Star Wars: The Phantom Menace prequel hit in March of that year and brought the still young Web to its knees. College basketball’s March Madness also provided a boost.
Dr. Leighton explained those eureka moments to ARC in a separate interview earlier this year:
I think in the early days the first eureka moment was when it went live and worked. One was in distributing the Star Wars trailer. We were the only entity on the Web that was able to do that. All the rest went down. Coincidentally that same day, on the first day of March Madness when ESPN.com went down in 1999, Infoseek (where it was hosted at the time) called us up and said, ‘we’ll do that trial now.’ We put them back online in 15 minutes and they were five-times faster than they had ever been, under enormous loads.
So those are eureka moments … and the whole world noticed both of those things. People wrote stories about them and it helped catapult our business. Hey, this really works. It is a totally different way of doing things. But it makes things faster and it makes things scale.
Like many Web-based companies in the late 1990s, Akamai filed its S-1 for its initial public offering at a relatively young age for the company. Akamai went public on October 29, 1999 at $26 a share. Akamai had shown initial technological and business success which put it ahead of many of the empty companies of that period that went public based on revenue and market projections that turned out to be faulty.
Akamai would not ride the high water mark for long.
The Passing Of Lewin And The Bubble
Akamai grew quickly in its first few years of existence, quickly rising to about 900 employees, mostly based in Cambridge.
And then came the gut punch.
Lewin passed away on September 11th, 2001 aboard American Airlines flight 11 that struck the first World Trade Center tower. As the story goes, Lewin—an Israeli-trained counterterrorism expert—was the first person to die in the September 11th attacks in an attempt to stop the hijackers.
“I learned when I came to the office. The phones were down, so you couldn’t communicate,” Dr. Leighton said of the day Lewin died. “When I was at home, I heard that there had been a crash. Initially I had heard incorrectly about what flights had been involved. When I came to the office, I immediately knew that it was Danny’s flight.”
The Web bubble burst between 2000 and 2002. Many of the companies that Akamai relied on as customers dried up and died or did not have the cash to pay for a content delivery network. Akamai’s stock price tumbled to less than $1.00 (a weekly low point of $0.71 on Sept. 30, 2002, per Yahoo Finance, hundreds of dollars down from its pre-bubble peak) and the company was forced to cut around 400 jobs.
“We had to swap out our customer base to get long-term customers, solid enterprises that would survive, the big Web companies. Lot of the fly-by-night Web companies that had become our customers died. So you had to replace that,” Dr. Leighton said. “And revenue was shrinking during that time because our customers were going broke and the ones that weren’t had to be more economical in what they were spending.”
Akamai survived the bubble through astute, if painful, management under then-CEO (and now chairman) George Conrades. Akamai reached free cash flow in 2003 and the break-even point in 2004. Conrades gave way to Paul Sagan as CEO in 2005 and Sagan served until 2013 when Dr. Leighton assumed the mantle (he had served as Chief Scientist previously).
“You have the bubble. During that phase everything is free, the food trucks come twice a day, blah blah blah,” Dr. Leighton said. “Then there is the crash. Now you’re laying everybody off and that’s a nightmare. You are cutting back everywhere, which hurts. Then there is the break even and making it where you get past throwing everything overboard.”
Akamai now has close to 6,000 employees and did almost $2 billion in revenue in 2014 and has offices all around the world. Three other command centers just like the one in Cambridge are continuously live in Bangalore, Krakow and Santa Clara.
“Obviously when Danny dies that’s a huge blow and you have to manage through that,” Dr. Leighton said. “You’ve got to manage through conveying to everyone that there is a light at the end of the tunnel. It is not the proverbial train coming at you, that you are going to be OK and there’s a reason why. Otherwise, people would be demoralized and the company does die.”
The Combination Of Mathematics And Computer Science
The type of traffic that Akamai sees is almost unfathomable. One of the reasons that Akamai is such a fascinating company is that it takes a distributed network—the technology available today—and applies advanced mathematical concepts to route traffic and make sure the Internet is running.
“We’re getting about 43 million hits per second, 18 terabits of data per second,” said Akamai network operations manager Kevin Lamont during a tour of the command center on a Thursday morning in September. “We even have a research base on the South Pole. Anywhere there are eyeballs, we can put our machines close to them and make that experience better for the end users.”
Lamont noted those traffic levels were a little high for a Thursday morning, but generally within accepted normal levels. Akamai had just broken a traffic record on July 28 when it saw 31 terabits of data per second on its network. Akamai does not usually say who its customers are and what type of traffic they are pushing, but July 28th happened to be the day that Microsoft began pushing Windows 10 to computers across the world.
Akamai’s traffic is driven on a daily basis mostly by media companies—images, videos, articles—with large software updates and video game releases causing spikes throughout the year. Akamai is organized around four primary business units: video/media, Web performance, security and enterprise networking and apps.
“Media is the biggest,” Dr. Leighton said. “It is ballpark half our revenue. Web experience, over a third of revenue. Security, over 10% and the fastest growing now. The cloud or enterprise network is really just getting started with revenue. We have a large team and doing a lot of development but just starting on revenue.”
How exactly does Akamai handle the immense flow of traffic running on its network? The trick is a combination of cheap components spread across a geographically diverse network with traffic control algorithms to automatically identify and route traffic around problem or congested areas.
Fundamentally, Akamai’s business breaks down to one incredibly complex math problem.
“It is the algorithms and the distributed nature of those algorithms in the hardware and the software that enables us to have that scale and volume,” Dr. Leighton said.
The algorithms are really what set Akamai apart and one part of the company where the presence of Lewin is still felt.
Dr. Leighton, who still teaches at MIT every other autumn, explains the core components of Akamai’s system.
You know, it’s called theoretical computer science or the designs of algorithms whose scalability is N what ever N may be. Be it users, traffic, whatever gets really large. That’s a branch of the intersection of mathematics and computer science. Which is where my background is, where Danny’s background was. Developing algorithms in that sense has given us a tremendous advantage in trying to build and scale the company efficiently and to get great performance.
The design philosophy of lots of little, cheap, dumb components (dumb, said in a sense) when put together become highly intelligent and highly cost effective with no single points of failure. In fact, you really want to have no several points of failure. As opposed to a centralized system where you just try to have that one data center have so many 9s and it is very expensive.
So it is a different approach architecturally than you see most anywhere else. That is really important for us, I think.
The Internet’s Tide Keeps Rising
The growth of the Internet and the Web shows no signs of abating. More and more of the world is coming online—mostly through smartphones—and the quality and volume of video increases daily. The Internet of Things will provide billions more end points to the Internet, all creating data that must be sent from one location to another.
“We set [the] record probably every four to six months,” Lamont said. “The traffic is increasing so quickly with every new, big event coming up and we are routinely breaking this record.”
Dr. Leighton promises that Akamai can handle three orders of magnitude more traffic on its network. Capacity is not really an issue nor the fact billions of more end points will soon be connected to the Internet. Akamai is set up to handle both. When it comes to video, Dr. Leighton notes that it will have to work with the entire technology and media ecosystem to provide higher quality video to all kinds of devices, including smartphones and set top streaming boxes and so forth.
The plan is for Akamai to carve up its four key areas of concentration and improve the capabilities of each, drawing inspiration and resources between groups. For instance, as more enterprise apps start using the Web and HTTP to move data, Akamai can tap into the lessons from its Web performance group to provide speed and security to large companies.
The digitization of everything provides Akamai with a basically unlimited addressable market. And Dr. Leighton, of course, believes that the Internet would be better served if Akamai was the entity carrying all of that data.
“But it is a lot of traffic and that is what we built the platform to handle, deal with a lot of scale. Offload the key peering points on the Internet. When Akamai is carrying that traffic, it is better for the whole world,” Dr. Leighton said.