No Better Time

  As VP of Strategy and Corporate Development, Seelig’s focus was Akamai’s infrastructure, a complex, highly sophisticated network of servers that remained in constant contact with each other, using their own unique language of Linux, a computer operating system. Seelig described the task of managing the installment of the servers—pre-programmed with Akamai’s software—in data centers worldwide as a logistical nightmare. An international shipping company was in charge of delivering the servers, along with the cables, switches, and instructions, to their physical destinations. Once the servers were on location, Akamai’s partners at the data centers installed them. Still, Akamai constantly experienced snafus related to shipping damages or extreme weather. In addition, as Akamai rapidly added servers around the world, its connection to servers in the most distant countries was sometimes so poor that Seelig said they had trouble confirming it.

  As chief scientist, Tom Leighton worked mainly out of Cambridge during Akamai’s first year. Occasionally, he traveled for sales calls or investor meetings, but the focus of his job was the technology—often referred to as Akamai’s “secret sauce.” As Leighton explained, “It was more of an inward role. I spent the bulk of my time improving our architecture.” But with a small staff and a never-ending list of responsibilities, Leighton—like everyone at Akamai—had to do a little bit of everything. “For a while, I was our customer service department,” said Leighton, who took many late-night calls when a connection unexplainably went down. “In those days, it was stressful because the service wasn’t really [fully] ‘baked,’ and our description of it was often far grander than what it really was,” admitted Leighton. “Every once in a while, I’d hear Danny promise something and think, ‘Uh oh, now we’ve gotta deliver on that, and it was a challenge.’”

  Akamai had a list of most wanted customers, and Disney was close to the top. But it was a tough culture to crack: a long established corporate giant with a Byzantine internal structure. Moreover, in 1998, Disney bought Starwave Corp./Infoseek, a Seattle-based software and Web company led by billionaire Paul Allen, who co-founded Microsoft. Starwave set the standard for much of the commercial Web explosion of the decade. Having recently merged with Infoseek, Starwave was developing major sites like ESPN.com and Mr.Showbiz.com. Disney’s acquisition of the Starwave/Infoseek conglomerate formed a new giant, Walt Disney Internet Group. Randall Kaplan made one of the first connections to the company through Jake Weinbaum, then chairman of Disney’s Internet Group.

  Weinbaum directed him to contact Randy Dragon, an engineer who managed all the online operations for Disney Interactive Studios in Los Angeles. Disney hired Dragon from Columbia Pictures in 1985. But it wasn’t until 1999 that the company started Disney.com under the leadership of Chairman Michael Eisner. Eisner’s vision for his company’s online presence was grand, and because of this he gave Dragon’s fifteen-person staff a few years’ free rein to build the site. As a corporate behemoth, Disney had first turned to top leaders in technology like ATT’s Bell Labs to help the company find the ideal hosting service. “But the big guys didn’t have a clue about infrastructure,” Dragon noted. “And neither did we.” With the desire to launch a Web site rich with media at a time of dial-up Internet connections, Disney was struggling. The acquisition of Starwave didn’t help—the company had its own engineers running ESPN’s blockbuster site, but Dragon didn’t see long-term promise in their infrastructure either. Toward the end of 1998, Akamai showed up with a solution. Dragon recalled one of their first meetings: “With their white socks and buttoned-up look, they were a little strange to us—they were so academic,” Dragon said. “But the idea they presented was so impressive, and it was clear they’d really thought it through. They weren’t trying to put the hard sell on us, they were really trying to create a solution, and it was amazing.”

  Dragon agreed to sign Disney on as one of Akamai’s first trial customers. Sometime in February 1999, Dragon called Akamai saying he felt the FreeFlow trial had gone really well, and that he wanted to sign up with Akamai, but needed to review the company’s standard Service Level Agreement (SLA). John Sconyers in sales took the call, and agreed to send one along that day. He hung up and realized they didn’t yet have any such agreement. He called on Leighton, who, with the help of several engineers and Akamai’s General Counsel Robert Ball, spent two hours drafting language that would hold Akamai accountable to customers for superior performance. Akamai still wasn’t generating any revenue, but they now had the promise of it. “We didn’t really have any business yet,” Leighton said. “We had one object deep on Disney’s site where no one would go. We were hitting it every few minutes ourselves just to see if it was still there.” Akamai needed to prove itself. The trial tests were running better than expected, but the company was still selling a vision; the promise of a service so powerful it could keep the most popular Web sites online during an unprecedented crush of traffic. For Akamai to fulfill that promise in real time, live on the Internet, they needed to test their software during a naturally occurring, high-traffic surge. As good luck would have it, two traffic surges happened soon thereafter. And by sheer coincidence, they took place on the very same day: March 11, 1999.

  It was the first day of “March Madness,” the annual NCAA basketball tournament. It’s one of the most-watched sporting events of the year, one that attracts millions of fans who tune in every day for three weeks to watch their favorite teams duke it out for the championship title. By the mid-1990s, at least two companies were broadcasting the tournament live on the Internet: Infoseek (which hosted ESPN’s Web site) and SportsLine USA. Both were experiencing historic amounts of traffic, with page hits exceeding 160,000 a minute.{36} Around lunchtime that day, someone from Infoseek called Akamai with two urgent questions: How could they sign up for FreeFlow? And could they do so immediately? Leighton was shocked. Akamai had been in aggressive pursuit of Infoseek, one of the most popular search engines on the Web, but had been rebuffed. Within fifteen minutes of the call, though, Akamai was delivering a portion of Infoseek’s traffic. The phone rang again. It was another call from Infoseek, with another pressing request: Could Akamai handle a load of 2,000 objects per second? “We had no choice but to say ‘Sure, we can do that,’” said Leighton. Within a few minutes, Akamai was successfully delivering three thousand objects per second for Infoseek. Leighton said they didn’t know then that the company was hosting the NCAA tournament on ESPN’s Web site, which had crashed almost as soon as the games began. With Akamai’s help, Infoseek was back online and five times faster than they’d ever been before.

  That night, Akamai was hosting a small gathering of investors for the screening of a trailer for the fourth, highly anticipated Star Wars film, Episode I: The Phantom Menace, which was premiering in theaters nationwide a week later. Earlier in the month, someone at Paramount’s Entertainment Tonight show had contacted Akamai asking if the company could deliver the trailer for its Web site on the night of its release. An agreement was signed. Unbeknownst to anyone at Akamai, Star Wars producer and director George Lucas had struck a deal with Steve Jobs, interim CEO at Apple, for a blockbuster release of the trailer on Apple.com and Lucas’s Starwars.com using Apple’s new QuickTime III video player technology.{37} Around 9:00 p.m. that night, a room full of Akamai employees and investors tuned in to the live stream of the trailer on the FreeFlow-enabled Web site of Entertainment Tonight, which went off without a hitch.

  Shortly after the screening, someone at Akamai burst in the room with breaking news: Apple’s Web site had crashed! So had a handful of other Web sites as they attempted to stream bootlegged copies of the Star Wars trailer to more than twenty million viewers worldwide. The only sites that remained live, streaming the trailer without any outages or delays, were Paramount.com and Entertainment Tonight. Akamai’s FreeFlow handled up to 3,000 hits per second for the two sites—250 million in total—and the system never exceeded even 1 percent of its capacity. In fact, as the download frenzy overwhelmed other sites, Akamai p
icked up the slack. Before long, Akamai became the exclusive distributor of all Phantom Menace QuickTimes, serving both Starwars.com and Apple.com.

  It was another victory for Akamai, one so significant that news of it quickly made its way through the industry, allowing the company to approach more choice customers with greater credibility. The first of these names was a crazy-sounding business out of Stanford called Yahoo.

  Yahoo burst onto the scene in 1994 when David Filo and Jerry Yang, both PhD candidates in Electrical Engineering at Stanford University, created an online list of their favorite Web sites. When the list grew so big it became unwieldy, they broke it down into categories, then subcategories. What started out as “Jerry and David’s Guide to the World Wide Web,” became Yahoo, one of the most popular search engines on the Web. In just a few months, Filo and Yang’s pet project had reached a peak of one million hits a day. The pair incorporated in March of 1995 with an initial investment of $2 million from the California venture firm Sequoia Capital. By the time Akamai launched, Yahoo was a household name. In the dot-com boom, a working relationship with Yahoo held so much clout it could put an otherwise struggling startup firmly on the map. And Yahoo knew this. “The environment was such that any idea [for a dot-com startup] had four or five people [working on] it at the same time, all of them backed by big investors, and the first thing they’d say is, ‘Go see AOL and Yahoo and get them,’” explained Farzod Nazem, an angel investor who served as Yahoo’s first chief technology officer (CTO). “I say this with the ultimate humility: we were kingmakers.”

  But even for Akamai, a company built on the idea of scale, Yahoo was a hard sell for one main reason: it was growing so fast that no one could catch up. By necessity, Yahoo was mostly self-sufficient, even when it came to managing its peak congestion. “The scope of Yahoo was so large at the time that there was literally no vendor that could handle our traffic,” said Nazem. “But we had a lot of small objects and a lot of traffic, and we needed to be able to serve these things without melting down.” Using a combination of hardware and software as well as partnerships with peering points—two networks who agree to accept and forward each other’s data packets—Yahoo was able to shoulder a heavy load. But the more Yahoo grew, the more urgent was its need to outsource.

  At the time, you had to know someone who knew someone at Yahoo to get a foot in the door. Fortunately for Akamai, that someone was Michael Moritz at Sequoia Capital, who helped Akamai secure a meeting with founder David Filo in 1998, but nothing came out of it. Lewin, however, wouldn’t take no for an answer. Sometime at the start of 1999, he and Sagan flew out to Yahoo’s Santa Clara, California offices to meet with a group including Filo and Nazem. Nazem reckoned that he and Filo might have seen representatives from every significant startup of the decade come through their offices, and smart, driven entrepreneurs like Lewin led most of those startups. But Lewin, they agreed, made one of the few lasting impressions from the whirlwind year that was 1999. “We were impressed with Akamai in general, and I think a lot of that had to do with Danny,” remembered Filo. “Most of the time, when you’re dealing with someone trying to sell something, you’re talking to a salesperson. But we were talking to Danny and the other founders, and they clearly had the strongest technical team out there.” Nazem concurred, and they decided to strike a deal with Akamai.

  Paul Sagan clearly recalled the moment Filo asked him and Lewin for the documentation he needed to become an official test customer. When they handed the stack of paperwork over to Filo, he took one look at it and asked if he had to complete it all to “Akamaize.” “Of course, we wanted to burst out loud cheering and kiss him,” said Sagan. But we remained very poker-faced and mumbled something about how we could certainly work something out to waive some of the requirements.” Sagan and Lewin assumed that, like most of their big-name customers, Filo would want to test the FreeFlow technology on one hidden pixel buried deep on Yahoo’s site. Instead, Filo surprised them with a request to stream the Visa logo, which the company displayed on every one of its home pages. “We told him something like, ‘OK, I suppose we can do that,’” Sagan said.

  Sagan and Lewin told Filo they’d have someone at Akamai immediately send over a content code he could use to retag the Visa logo. When he’d done this, Filo just needed to email it back to Akamai’s offices in Cambridge, and they’d immediately start serving the image. On the way out of the building, Lewin called the student who was in charge of the programming team that day. “He told him, ‘You’re going to get a message from David Filo—the real David Filo—so you need to make sure you don’t fuck this up!’” Within minutes, they got a return call from the student saying that Filo had successfully Akamaized the logo. Yahoo was officially a client. Lewin called Leighton with the news. “It was mind-blowing for us,” declared Leighton. “At the time, Yahoo was the who’s who of the Internet, and suddenly we were [delivering a portion of content] on their homepage.”

  Sagan and Lewin left Yahoo’s offices in their rental car, which had, by chance, been upgraded to a Mustang convertible. On their way up Highway 101, they opened the roof and shouted at the top of their lungs: “Ya-hooooooooooo!”

  If anything had the potential to demonstrate the limitations of the Internet under stress, it was breaking news. “Some of the largest and most unpredictable hot spots were in the news business,” said Sagan, who experienced the problem firsthand when he served as president of New Media at Time, Inc. And when news sites crashed, the cost to the organization was immeasurable. “Reputations in news could be made or broken on the day of a big story,” Sagan said, noting that a banner day of breaking news coverage could result in a ten-year run of dominance. The same was soon true, he explained, of news online: “It became essential to win stories on the Web.” Akamai was perfectly poised to help news organizations that wanted to win more online followers. The biggest and most popular of these was CNN.com.

  At the end of August 1995, CNN Interactive launched CNN.com. It rapidly became one of the busiest sites on the web, and within two years had expanded to include CNNFN (Financial Network), and CNNSI (Sports Illustrated). Sam Gassel, then the chief systems engineer for CNN Internet Technologies, had served as the architect of CNN’s web system since its launch. Even before joining CNN he had been building web servers as a systems administrator for the University of Chicago.

  One of Gassel’s greatest challenges was managing the flash crowds generated by breaking news, which he and his team of experienced engineers were working diligently to tackle. When traffic peaked, they turned to a large arsenal of technological weapons and strategies including the reallocation of a server from one of CNN’s less trafficked Web sites to one struggling or stripping content down to the bare minimum by removing photographs and logos that occupied bandwidth. As early as 1997 they had also begun to experiment with caching technologies.

  Gassel first met with Jonathan Seelig and John Sconyers in the fall of 1998. Recognizing the Akamai concept as a logical step forward, he was intrigued, but skeptical. “I told them if you can build this and run it—and we have our doubts—then get back to us.” The dialogue continued and after a few more calls with Sconyers, Seelig and Lewin, Gassel agreed to test an invisible image on one of CNN’s sites as part of Akamai’s test trial.

  A few months later, Lewin and Sagan took a trip to Atlanta determined to score CNN as a customer. Lewin was armed with what he believed to be a compelling point in his sales pitch: CNN could one day see a news event so huge that no amount of servers or bandwidth could handle it. The Internet users of the world would turn to CNN, which billed itself “the most trusted name in news,” only to find blank or skeletal Web pages.

  Gassel clearly recalls his first encounter with Lewin: “He was a bundle of energy,” he said. “And he was awestruck when he saw both the newsroom and all our racks of servers.” Sagan, who knew Gassel from when he [Sagan] had worked at Time Inc.’s Pathfinder likened that first face-to-face conversation between Lewin and Gassel to a
n intellectual ping-pong match, with each of them lobbing high-level concepts at the other.

  Gassel also recalls a later visit that was interrupted by breaking news. Watching and listening to the CNN engineers methodically reconfiguring their systems to redirect traffic and avert a crash, Lewin remarked: “You guys sound exactly like Israeli tank commanders!” Gassel added: “He reassured me that this was a compliment.”

  Other news organizations followed, including The Washington Post. For years, Eric Schvimmer, director of operations for the newspaper’s Web site, had been trying to combat flash crowds using every possible method. But he and his staff were failing, Schvimmer said, when it came to huge news events like the 1996 presidential election. By 1999, though, Schvimmer admitted, the problem “was keeping me up at night.” “We were scaling more and more hardware, but we just couldn’t keep up on the server side.” Schvimmer had met with Lewin and Leighton a few months earlier and was impressed with what they proposed. But he also remained skeptical. Once he heard Akamai was serving traffic for CNN, however, Schvimmer petitioned the Post to invest in a service agreement with Akamai. “It was a no-brainer to me,” he said. “We had this opportunity to get the very same service that the big boys had.” By February 1999, the newspaper was testing content with Akamai, and its Web site was functioning better than Schvimmer could have ever imagined.

  There’s good publicity and bad publicity. And then there’s the kind of publicity no amount of money can buy. In the late 1990s, a glowing feature on an up-and-coming company posted on The Motley Fool’s Web site could send Wall Street into a state of frenzy. Founded in 1993 by brothers David and Tom Gardner, Motley Fool’s free Web site, Fool.com, had grown from a readership of just sixty into one of the most consulted online financial forums, one popular for its candid, humorous, and skeptical financial news and advice. (The company took its name from an Elizabethan drama, where only the court jester, or fool, could tell the king the truth without getting his head lopped off.)