Danny Lewin

On March 11, 1999, two of the heaviest loads the young World Wide Web had yet seen arrived on the same day. ESPN's website was covering the opening day of the NCAA basketball tournament, March Madness, with fans refreshing scores from offices across America. The same day, Paramount's Entertainment Tonight site posted the second trailer for Star Wars: Episode I – The Phantom Menace, the first new Star Wars film in sixteen years. Many of the downloads were files of about 25 megabytes, enormous for a time when most people connected through a telephone modem.

Sites that tried to carry that kind of demand on their own servers tended to fail. The problem even had a name among computer scientists, the hot spot: when too many people requested the same thing from the same place, the machines holding it overheated in the figurative sense, and nobody got anything. In 1998 Congress's servers had been unable to keep up with demand for the Starr Report. That spring, several sites distributing Phantom Menace trailers went down.

Both ESPN and Entertainment Tonight were quietly testing the service of a startup in Cambridge, Massachusetts, that was barely seven months old. According to Wired, its network handled up to 3,000 requests per second for the two sites that day, some 250 million in total, and never used even 1 percent of its capacity. Soon afterwards the company became the exclusive distributor of the Phantom Menace QuickTime trailers on both the official Star Wars site and Apple's.

The company was Akamai Technologies. Its technology rested on a set of algorithms that a 28-year-old graduate student named Danny Lewin had begun developing as part of his master's thesis at MIT. Lewin was not a typical graduate student. He was a married father of two, a former officer in Israel's most elite commando unit, and a bodybuilder who could explain the mathematics of distributed caching while bouncing around a whiteboard. Over the next two and a half years he would help turn Akamai into one of the most celebrated companies of the dot-com boom, fight to keep it alive through the crash that followed, and die on the morning of September 11, 2001, aboard American Airlines Flight 11. He is widely regarded as the first person killed in the attacks.

Daniel Mark Lewin was born on May 14, 1970, in Denver, and grew up in the suburb of Englewood, the eldest of three sons. His father, Charles, was a psychiatrist and his mother, Peggy, a paediatrician. His biographer, Molly Knight Raskin, describes a household that took ideas seriously. Charles disliked what he saw as the materialism of American culture and tried to fill his sons' lives with better material; when he found the boys looking for a cartoon character on the back of a cereal box, he covered the box with an article from Scientific American. After that, puzzles in maths and science became a family tradition at the breakfast table. In 1979 Charles brought home an Apple II, and Danny, then nine, taught himself to program it.

When Danny was fourteen, Charles announced that the family was moving to Israel. He had become drawn to Zionism and wanted to make aliyah, the Jewish term for settling in the land of Israel. Nobody else in the family wanted to go. Raskin writes that Danny was furious, and that for about a year he barely spoke to his father. After three months in an absorption centre for new immigrants near Jerusalem, the family moved into a small apartment in the city's French Hill neighbourhood.

Danny struggled with Hebrew and with Israeli social codes, and he found school too easy to take seriously. He channelled his anger into the gym. At Samson's Gym in Jerusalem he trained until his muscles gave out, and by sixteen he could bench-press more than 300 pounds. He also taught himself mathematics, and took the American SAT for fun, scoring close to perfect. Gradually he came to think of himself as Israeli-American, with strong views about his adopted country's security.

It was at the gym that he heard stories from men who had served in the army's special forces. By the time he was called up at eighteen, he had decided that he would serve only in a top combat unit.

For a recent immigrant with no family history in the Israel Defense Forces, the ambition was presumptuous. Sayeret Matkal, the General Staff reconnaissance unit, is the most selective formation in the Israeli army, a counterterrorism and intelligence commando force often compared to the American Delta Force. Its selection process is designed to break most candidates. Lewin passed it and served in the unit for about four years, eventually becoming an officer. Accounts of his service, including a 2004 report based on the 9/11 Commission's findings, give his rank as captain. As part of his training, Raskin writes, he learned Arabic and studied how terrorists operated. To earn the unit's beret, he completed a march of 75 miles in 25 hours.

Friends who served with him remembered his physical intensity. His closest friend from Israel, Marco Greenberg, later called him the most formidable person he had ever met.

Near the end of his service, Lewin met Anne, a Belgian who had moved to Israel with her family as a teenager. They married in Jerusalem, and their first son, Eitan, was born about a year later. Lewin left the army to study at the Technion – Israel Institute of Technology in Haifa, the country's leading engineering university. While studying, he also worked full time: from May 1994 to May 1996 he was a research fellow and project leader at IBM's research laboratory in Haifa, responsible for developing and supporting a verification system called Genesys. He also served as a teaching assistant.

In 1995 he graduated summa cum laude with two degrees, a BA and a BS, and was named the Technion's outstanding student in computer engineering. Somewhere in those crowded years he found a book on parallel algorithms by F. Thomson Leighton, a professor of applied mathematics at MIT who ran its algorithms group. Lewin decided he wanted to work with its author.

He applied to ten of the leading computer science graduate programmes in the United States and, according to Raskin, was accepted by all of them. He chose MIT. His father was unhappy; to Charles, leaving Israel was a kind of betrayal, and he asked his son to promise to return after his doctorate. Lewin intended to. He hoped one day to teach at the Technion.

Lewin arrived at MIT's Laboratory for Computer Science in 1996 as a doctoral candidate. Anne, who was pregnant with their second son, Itamar, followed a month after giving birth. Money was tight. Raskin describes a small apartment, a diet of canned and boxed food, Lewin taking odd jobs such as sweeping the halls of their building, and a classmate recalling him agonising over whether they could afford to replace one of the boys' lunch boxes.

The lab was one of the most important places in the world to think about the Web. Tim Berners-Lee, who had invented the Web in 1989, had his office there, running the World Wide Web Consortium. In early 1995, according to Akamai's later account in its stock-market prospectus, Berners-Lee had put a challenge to his MIT colleagues: find a fundamentally better way to deliver content across the Internet. Leighton's algorithms group took it up. The hot spot problem was interesting in theory and urgent in practice. The graphs of nodes and connections that the group drew on whiteboards stood for real machines and real cables.

The Web's architecture made hot spots almost inevitable. A web address was tied to a particular set of servers in a particular place. A company could buy more machines, but they still sat at one end of a long chain of networks, and every user had to reach them through the same congested routes. Caching, which stored copies of popular files closer to users, helped, but it was hard to keep the copies fresh, and it broke down when servers came and went.

Lewin was older and more worldly than most of the students around him, and he stood out in other ways. Leighton later recalled that he rose above his classmates quickly. He wrote limericks and pinned them to the lab walls, played pranks, and worked without apparent need for sleep. One fellow student, Yevgeniy Dodis, told Raskin that he had the drive of a madman. Leighton took him on as a research and teaching assistant and agreed to supervise his master's thesis, and the two became close collaborators, sharing an office and sketching algorithms late into the night.

The idea that became Lewin's thesis was called consistent hashing. Hashing is a standard way of deciding where to put things: a formula turns each item's name into a number that points to one of a fixed set of locations. The trouble is that ordinary hashing depends on the number of locations. Add or remove a single server and the formula sends almost every item somewhere new, which, in a network where machines fail constantly, means chaos. Wired compared it to a card dealer whose system for dealing hands collapses as soon as a player joins or leaves the table.

Consistent hashing solved that problem. It assigned both items and servers to positions in the same abstract space and mapped each item to a nearby server, so that when one server dropped out, only the items it held needed to move, and they moved to a neighbour. The scheme spread load evenly across machines, tolerated failure and did not require every part of the network to have the same up-to-date view of the whole. In May 1997 the idea was presented at the ACM Symposium on Theory of Computing, the field's leading conference, in a paper by David Karger, Eric Lehman, Tom Leighton, Matthew Levine, Daniel Lewin and Rina Panigrahy. The paper suggested that consistent hashing might one day be useful well beyond Web caching, and it was right: the technique later became a standard building block in large distributed systems.

At first Lewin had doubts. Raskin reports that a faculty adviser dismissed the idea as insignificant, and that Lewin once told a research partner that consistent hashing was a pathetic idea, but that it was his idea. According to Raskin, a paper on the concept was rejected by one academic conference, whose reviewers did not think it had any hope of being useful. Leighton, by contrast, thought it was a gem, and he believed its simplicity was the point.

The group went further than a single algorithm. For most of their methods they proved that no alternative could do better; for the rest they showed that the underlying problems were inherently hard, so that no shortcut existed. They also became convinced that the larger the network grew, the better their approach would perform. In 1998 Lewin's thesis won an MIT award for the year's best master's thesis presentation.

The step from theory to company came partly through a friend. In the spring of 1997, Preetish Nijhawan, a student at MIT's Sloan School of Management, urged Lewin to turn his thesis into a business plan and enter the annual MIT $50K Entrepreneurship Competition. Lewin, Leighton and Nijhawan won the software category in a preliminary round. According to Raskin, they spent the $100 prize on beer.

The final took place in May 1998. There were 84 entries. The Akamai team, which by then included Jonathan Seelig, another Sloan student, finished fifth. Raskin writes that Lewin had been sure they would win and took the result hard.

The loss turned out to be useful. The team had learned what investors wanted to see, and the competition had put them in front of the right people: Battery Ventures, whose partner Todd Dagres became one of Lewin's early backers, discovered the company through it. Over the summer of 1998 a smaller team turned the algorithms into a working prototype and a better business plan. Lewin faced the hardest choice. Leighton could return to his professorship if the venture failed; Lewin had a young family, little money and an MIT stipend he would have to give up. He took the risk.

Akamai Technologies was founded in August 1998 by Lewin, Leighton, Seelig and Randall Kaplan, a young executive who had worked for the financier Eli Broad. Lewin became president. The name was his suggestion: the team found it by trying words in an online English–Hawaiian dictionary, and it means clever, or, more loosely, cool. In October 1998 MIT granted the company an exclusive licence to the relevant intellectual property in exchange for shares.

The first money came from family, friends and angel investors, among them Gil Friesen, the former president of A&M Records. Friesen, who later admitted he had no idea what Akamai actually did, invested half a million dollars after watching Lewin at a whiteboard; he told Raskin the experience reminded him of seeing great musicians before they were famous. Battery Ventures and Polaris Venture Partners followed, and by the end of 1998 the startup had raised more than $8 million. In January 1999 Paul Sagan, a former president of new media at Time Inc., joined as chief operating officer. In April George Conrades, a former senior IBM executive and chief executive of the network company BBN, became chairman and chief executive. In May Lewin handed over the title of president and became chief technology officer, the role he held for the rest of his life.

Akamai's early decisions shaped the industry that followed. The founders first imagined selling their algorithms as software that Internet service providers would install. They soon concluded that they should instead run their own network of servers, placed inside as many other companies' networks as possible, and sell guaranteed fast delivery as a service. A customer ran a small utility that rewrote the addresses of the heavy parts of its pages, such as images, video and software downloads. When a user requested one of those files, Akamai's name servers worked out where the user was on the Internet, which servers were nearby, and which were lightly loaded, and sent the request there. Consistent hashing spread the requests among the servers in each region. The customer kept control of its main page and its data, and the heavy traffic never reached its own machines.

Building that network quickly required salesmanship as much as mathematics, and Lewin turned out to be exceptional at both. Raskin's book collects stories of his persistence. When Microsoft executives told him they did not need Akamai, she writes, he refused to leave until security escorted him out, then returned the next day and waited in the lobby until an executive agreed to hear him again. In September 1999 Microsoft invested about $15 million in Akamai as part of a partnership to carry its Windows Media streaming technology; Cisco had invested about $49 million a month earlier.

His whiteboard performances became company legend. Dwight Gibbs, the chief technologist of the financial site the Motley Fool, recalled a room of more than twenty sceptical engineers slouching in their chairs as Lewin began. As he filled the board with equations and strategy, becoming more animated as he went, they leaned forward. When he ran out of whiteboard, he kept writing on the wall. Afterwards, Gibbs said, his team told him Lewin was scary smart and that the idea might actually work.

Lewin's approach to operational problems was similarly direct. Raskin recounts that when a hardware supplier said it could not build around a thousand servers in the four weeks Akamai needed, Lewin bought tool kits, flew to the supplier's plant in North Carolina with several engineers and helped assemble the machines himself. The early network was held together with improvisation; one batch of servers arrived with badly fitted screws that came loose in data centres.

The customer list grew quickly. Akamai served its first live traffic in February 1999, a single tracking pixel on the Disney website, and launched commercial service that April, announcing Yahoo! as a charter customer. The same month it began a strategic alliance with Apple to deliver QuickTime streaming video. In 1999 Apple and Yahoo! accounted for 22 percent and 13 percent of Akamai's revenue. By October the company had 1,475 servers in 24 countries across 55 networks, and 193 full-time employees.

It also had a competitor. In September 1998 a California company, Sandpiper Networks, had launched a similar service. Lewin's response, as friends and colleagues described it to Raskin, was to treat the market as a contest he intended to win. Money interested him mainly as a way to give his family a comfortable life; beating rivals interested him a great deal.

Akamai went public on the Nasdaq on October 29, 1999, fourteen months after its founding. Demand was so strong that the expected price range was raised from $16–$17 per share to $21–$23, and the offering was finally priced above even that, at $26. The company also increased the number of shares on offer to nine million, raising $234 million.

The first public trade was at $110. The shares reached $166 during the day and closed at $145.19, a gain of 458 percent over the offering price and, at the time, the fourth-largest first-day rise for an American stock. The following Monday the company's market value approached $16 billion. Its leaders declined the traditional invitation to ring the Nasdaq opening bell and spent the day working in Cambridge.

On paper, Lewin, then 29, and Leighton were now among the richest people in American technology. Lewin's purchases, Raskin writes, were modest by the standards of the era: a house in Brookline and a few motorcycles to ride with friends. He planned to re-enrol at MIT and finish his doctorate. Sagan joked to Wired that Lewin was free to leave and publish his thesis, but that the company would then have to kill him, because he knew too much.

The stock kept climbing. On the last trading day of 1999 it closed at about $327.

The collapse began a few months later. As the Nasdaq fell through 2000, investors lost patience with loss-making Internet companies, and Akamai's customers, many of them dot-coms themselves, started going bankrupt. By July 2001 Forbes reported that Akamai had lost 150 dot-com customers in a single quarter, more than a tenth of its base, and that its shares had fallen to $7.60. By the summer they were around $5.

The company's bet was a new service, EdgeSuite, designed to deliver far more of a website, including whole pages, from servers at the edge of the network. At the company's 2001 sales kickoff in Las Vegas, Lewin announced it with a joke about its unlovely name, then acknowledged the company's losses and made his case for staying ambitious.

In April 2001 Akamai made its first layoffs, cutting more than 180 jobs, about 14 percent of its workforce, and warned that revenue would fall well short of expectations. Lewin pushed his managers harder. At 10:30 one night in May he sent his project management team a four-page email on leadership. He told them that there were many things a great leader did, but only three that mattered most: lead by example, suffer together, and hold people accountable while removing those who did not perform.

The pressure took a toll at home too. Raskin, drawing on interviews with his friends and colleagues, reports that Lewin and Anne separated for a time during 2001, and that by September they had reconciled. That April, Forbes had estimated his net worth at $285.9 million, a fraction of his paper wealth at the peak.

In the first week of September 2001, Charles and Peggy Lewin visited Boston. It was Charles's first visit since Danny's family had left Israel five years earlier. They toured Akamai and walked around Cambridge with their son. Friends and colleagues who saw Lewin that week remembered him as unusually settled.

On Monday, September 10, Lewin led a meeting of more than a dozen employees that lasted over eight hours, laying out a new direction for the company. Colleagues urged him to cancel a trip to Los Angeles the next day so that he could stay to manage the restructuring; he considered it and told them they would be fine. That evening he and Leighton sat down to decide which of the company's roughly 1,500 employees would lose their jobs in a second, larger round of cuts, about 500 people. Leighton later told Raskin that it was a horrible night: they had recruited these people, who were also their friends. The two men worked until two or three in the morning.

A few hours later Lewin drove to Logan International Airport in Boston for American Airlines Flight 11, the scheduled nonstop to Los Angeles. He had flown the route so often that, by Raskin's account, he knew the crew by name. He took his seat, 9B in business class, and spent the last minutes before departure on the phone with Akamai's in-house lawyer about a pending deal, until he was told to switch off his phone. The Boeing 767 took off at 7:59 a.m. with 81 passengers and 11 crew.

Five of the passengers were hijackers. Mohamed Atta and Abdulaziz al-Omari were seated in business class in the row directly in front of Lewin. Satam al-Suqami was in 10B, directly behind him.

What happened next is known mainly from two flight attendants in the coach cabin, Betty Ong and Madeline Amy Sweeney, who telephoned American Airlines on the ground and calmly relayed what they could see and hear. The hijacking began at about 8:14. At 8:19 Ong reported that the cockpit was not answering and that someone had been stabbed in business class. Sweeney reported that two flight attendants had been stabbed and that a passenger in business class had had his throat cut, apparently by a hijacker. The passenger, she said, was in 9B.

Early accounts were confused. An internal Federal Aviation Administration memo written shortly after the attacks stated that the passenger in 9B had been shot; FAA officials later said the reference to a gun was a mistake, and the final version dropped it. The 9/11 Commission, which interviewed those who spoke with the flight attendants, concluded that Lewin had probably been stabbed by al-Suqami. It suggested that Lewin, with his Israeli military training, may have tried to stop Atta or al-Omari in front of him, without realising that another hijacker was sitting behind him. Nobody can know exactly what he did in those minutes. Friends who knew him never doubted that he fought.

At 8:46 a.m. Flight 11 struck the North Tower of the World Trade Center, killing everyone on board.

Within minutes of the crash, engineers in Akamai's network operations centre in Cambridge noticed an unusual surge in traffic. As the scale of the attacks became clear, much of the world tried to reach the same few news sites at once. Raskin calls it the Web's equivalent of a hundred-year flood.

Akamai's chief executive and chief financial officer were in California and hard to reach as phone networks jammed, so Sagan took charge. He had to tell the company that its co-founder had been on the first plane, and to decide whether it should keep working. Many employees, he recalled a decade later, felt it would be wrong to carry on.

Akamai's network carried traffic that day for news organisations, including ABC and The Washington Post, and for agencies such as the Red Cross and the FBI. CNN, which had stopped using Akamai earlier that year, struggled to keep its site online as its traffic kept multiplying. According to Raskin, around noon it asked Akamai for help, and by 1:30 p.m. Akamai was serving the images on CNN.com. The technology Lewin had designed for the unpredictable spikes of news was given its hardest test on the day of his death.

It did not save the company from the bust. Akamai made the cuts Lewin and Leighton had planned, and more. For 2001 it reported a net loss of about $2.4 billion, much of it from writing down the value of acquisitions made at boom-era prices. In October 2002 its shares closed at 67 cents. Leighton later said the loss of Lewin had come on top of everything else: a customer base going broke, the bubble bursting, and money being lost hand over fist. But Akamai survived. It shifted towards large corporate and government customers, reached a profit on an adjusted basis at the end of 2003, and reported its first full year of net income in 2004, earning $34.4 million on revenue of $210 million. It later became one of the largest providers of content delivery and cybersecurity services in the world. Leighton became its chief executive in 2013.

Lewin was buried at Sharon Memorial Park outside Boston. In October 2002 the city of Cambridge named the intersection of Main and Vassar Streets, near MIT and Akamai's offices, Danny Lewin Square; his son Eitan unveiled the sign. The best student paper award at the Symposium on Theory of Computing, the conference where consistent hashing was first presented, now bears his name. Akamai holds an annual memorial for him beside its headquarters, where an apple tree was planted in his honour. In 2012 it bought the Israeli company Cotendo, fulfilling his long-held wish for the company to have a presence in Israel.

When Leighton gave a lecture in Lewin's memory at the Technion that year, a student asked how Akamai had managed to succeed in such turbulent times. Leighton, Raskin writes, answered almost immediately: one word, really. Danny.