Tubes: A Journey to the Center of the Internet
Page 18
Simon Cooper was Tata’s Englishman, with the job of making the company’s investment pay off. Internet traffic has grown continually in the last decade, but prices have fallen just as fast. Tata planned to buck the trend by finding the places in the world with latent potential. Its strategy was to be the telecommunications network that finally linked the “global south,” the poorer—and less connected—regions of the world, especially Africa and South Asia. Cooper spent his time trying to decide what countries to plug in next. Recently, he’d begun an ambitious building program to supplement the original Tyco network with even more cables—stringing them around the earth like lights around a Christmas tree.
In New Jersey, I waited a few minutes in the office kitchen, watching a group of Indian engineers make tea. Then, just before ten o’clock, I was invited into a conference room dominated by three giant flat-screen televisions, lined end to end facing a long table. Cooper was sitting inside the middle screen. He was in his early forties, with a shiny pate and a cheerful grin, looking a little worn out, alone in a room in Singapore late at night, coming to me via Tata’s high-end video-conferencing link. We’d spoken once before. That time, Cooper was in an airport lounge in Dubai at midnight. He seemed to be constantly roaming, physically and mentally, as if he were the human incarnation of the network itself. I suppose the fact that I was talking to a TV had something to do with it, but I couldn’t shake the notion of Cooper as a man inside the Internet. In a business filled with obfuscation, he was good humored and direct. I knew why: Tata was eager to compete with the AT&Ts and Verizons of the world, which meant improving their name recognition in the United States—and inviting over any journalists who asked.
“We’ve done the belt around the world and now we’re reaching up and down a little bit,” Cooper said nonchalantly, talking about the planet as if it were his lawn. Tata had extended its cable between the United States and Japan with a new link to Singapore and then onward to Chennai. Then from Mumbai another Tata cable passed through the Suez to Marseille. From there, the routes went overland to London, and finally connected to the original transatlantic cable that connected Bristol, England, to New Jersey. Cooper made it sound like no big deal, but he’d built a beam of light around the world.
To go “up and down,” Tata bought a stake in SEACOM, the new cable to South Africa, as well as another new cable down Africa’s western coast, intended to break SAT-3’s grip. And they were making a move into the Persian Gulf, planning a new cable that would connect Mumbai to Fujairah on the Emirates’ eastern shore, and then around the Strait of Hormuz to Qatar, Bahrain, Oman, and Saudi Arabia. The cable would go from port to port around the gulf like a packet ship.
“You get a number of benefits from being global,” Cooper said from inside the screen, chopping at the desk on the other side of the world. “We’re connected to thirty-five of the biggest Internet exchanges around the world, so you can get to DE-CIX or AMS-IX or London, whether it’s the last mile, or the last three thousand kilometers. And we get to talk about our global restoration, our round-the-world capability.” In other words, Tata could promise that if its path from Tokyo to California were somehow obstructed—by an earthquake, say—they’d happily send your bits around the other way. It reminded me of Singapore Airlines’ two daily flights from New York to Singapore: one goes east and one goes west. But only with the Internet do we treat the scale of the planet so casually—and only then because we have physical links like these.
For Tata, it was all an effort toward connecting the unconnected places—and thereby getting away from the falling prices on the generally overserved routes across the Atlantic and Pacific. “Look at Kenya,” Cooper said. “Last August it had only satellite. Suddenly it’s as well served as most other coastlines around the world, with the exception of hot spots like Hong Kong that have ten or twelve cables. But it’s gone from zero to three cables in eighteen months. That makes it part of the global network. Not every customer wants a link from Kenya to London, but once you can do it, and do it consistently and do it well, people begin to think about things like call centers, which are constantly hunting for the place with the lowest cost services. The demand springs up.”
Undersea cables link people—in rich nations, first—but the earth itself always stands in the way. To determine the route of an undersea cable requires navigating a maze of economics, geopolitics, and topography. For example, the curvature of the planet makes the shortest distance between Japan and the United States a northern arc paralleling the coast of Alaska and landing near Seattle. But Los Angeles has traditionally been the bigger producer and consumer of bandwidth, exerting a southward tug on earlier cables. With TGN-Pacific, Tyco solved the problem the expensive way, by building both.
Further complicating the geography is the demand for low “latency,” the networking term for how long it takes information to travel across the cable. Latency used to be a concern only for the strictly telephone people, eager to avoid an unnatural delay in conversations. But more recently it’s become an obsession of the financial industry, to serve the needs of high-speed automated trading, where computers arbitrage based on knowing the market news an extra millisecond in advance. Since the speed of light through a cable is consistent, the difference is entirely in the length of its path. Tata’s route from Singapore to Japan is more direct than its competitors’, which also gives it the fastest travel times all the way to India. But Tata’s transatlantic cable is frustratingly slow. Tyco originally connected it to a landing station in New Jersey, close to its corporate headquarters. But compared with the transatlantic cables that landed on Long Island, by the time a bit went down the coast and back up to the city, the route effectively made London and New York two hundred more miles apart. At the time no one thought it mattered. “Now I get beaten up in meetings because there’s one millisecond extra compared to our competitors,” Cooper said, rubbing his brow. The first new transatlantic cable in a decade will be laid in 2012 by a small company called Hibernia-Atlantic. They designed it from scratch to be the fastest.
The micro geography matters just as much. Specialized ships conduct surveys of the ocean bottom, carefully plotting routes over and around underwater mountains—like grading a railroad, but without the option to dig any tunnels. The paths carefully avoid major shipping lanes, to limit the risk of damage from dragging anchors. Because if a cable does fail, a repair ship is dispatched to lift both ends to the surface using grappling hooks and fuse the ends back together—an expensive, slow process. Occasionally, the situation becomes more dramatic.
All but a few cables between Japan and the rest of Asia pass through the Luzon Strait, south of Taiwan. Looking at the map it’s easy to see why: routing south around the Philippines would add too much mileage—and therefore cost and latency. But the Formosa Strait between Taiwan and mainland China is dangerously shallow, putting any cable at risk of being struck by fishermen. What’s left is the Bashi Channel in the Luzon Strait, which with depths of up to four thousand meters seemed the perfect cable highway.
Perfect, that is, until Boxing Day 2006, when just after eight in the evening local time a 7.1 magnitude earthquake struck south of Taiwan, causing major underwater landslides that severed seven of the nine cables passing through the strait, some in multiple places. More than six hundred gigabits of capacity were knocked offline, and Taiwan, Hong Kong, China, and most of South Asia were temporarily disconnected from the global Internet. Taiwan’s Chunghwa Telecom reported that 98 percent of its capacity with Malaysia, Singapore, Thailand, and Hong Kong was offline. The big networks scrambled to reroute their traffic on the working cables, or send it around the world the other way. But trading of the Korean won was temporarily halted, an Internet service provider in the United States noticed a sharp decrease in Asian-born spam, and a provider in Hong Kong publicly apologized for YouTube’s slow speed—a full week later. It was two months before things were back to normal. And the name “Luzon” still brings shudders to network engin
eers.
At the logical level, the Internet is self-healing. Routers automatically seek out the best routes among themselves. But that works only if there are routes to be found. At the level of physical cables, rerouting traffic means creating a new physical path, stringing a fresh yellow patch cable from the cage of one network to the cage of another—maybe in Equinix’s facility in Tokyo, or at the Palo Alto Internet Exchange, or inside One Wilshire in Los Angeles, all of which are places where the major transpacific networks have meeting points. But otherwise network owners are faced with the excruciatingly analog task of snatching up cables from the ocean floor with steel hooks. After Luzon, Tata had three ships in the area for nearly three months, picking up cables, splicing them together, lowering them back down, and then moving on to the next break. So when Tata planned a new cable in the region—the first postearthquake—Cooper thought twice about the route. “We went as far south as we could, which maybe isn’t the optimal route from Singapore to Japan, but if there’s an earthquake in the old place, we won’t be affected, and if there’s an earthquake near us, the other networks will generally stay up,” he told me, sitting up in his chair in Singapore, which matched mine in New Jersey. “You make these tactical decisions.” And then you turn the decisions back to the economics. Vietnam has eighty million people and poor connectivity. “Maybe they’d be interested in a new cable?” Cooper ventured. I tried to imagine what that would be like, a new cable hauled up on a white sand beach in Vietnam. Of all the moments of the Internet’s construction, it struck me as the most dramatic—the literal plugging in of a continent. I asked Cooper if Tata might have a new cable landing anytime soon. With enough warning, and if they didn’t mind, I’d try to be there to see it.
“Actually we do have a landing coming up,” he said from inside the TV.
“Where!?” I shouted. Then I got worried. What if it was on the other side of the world, maybe in Guam (a big cable hub) or Vietnam? Or what if it was somewhere not entirely conducive to visiting journalists watching critical infrastructure, like Bahrain, or Somalia? This might not be so easy. But Cooper was cool. “It depends on the weather,” he said. “We’ll let you know.”
In the meantime, I set out for the spiritual home of undersea cables. If the Internet’s newest links tended to settle in the corners of the map, the old ones concentrated in more familiar places, and in one place above all others: a small cove called Porthcurno, in Cornwall, near the western tip of England, just a few miles from Land’s End. Throughout the entire 150-year history of underwater communications cables, Porthcurno has been an important landing spot, but also a training ground—the cable world’s Oxford and Cambridge. Looking at a map, I could easily see why. The geography hadn’t changed. Land’s End was still the westernmost point of England; England was still a hub for the world. According to TeleGeography, the busiest intercontinental route is between New York and London—primarily from 60 Hudson to Telehouse North. Several of the most important physical paths passed through Porthcurno.
But visiting a cable landing station wasn’t as easy as getting inside the big urban hubs. The Docklands, Ashburn, and others had a constant stream of visitors. Security was tight, but there was a sense of them as inherently shared places, nearly public ones. But the cable landing stations were quietly hidden away, and they rarely received visitors. But Global Crossing, then the operator of a major transatlantic cable known as Atlantic Crossing-1, finally responded to my entreaties—perhaps officials were pleased I was paying attention to something other than the company’s spectacular 2002 bankruptcy. My press contact only asked that I have a chat with her director of security, who would in turn “notify his government contacts” of my plans. Ah yes, those plans: to visit the Internet.
A little while later I was boarding a Penzance-bound train at London’s Paddington Station. The iron arches of its rail shed were the perfect send-off. Paddington was designed by Isambard Kingdom Brunel, the greatest of Victorian engineers, who also laid the route and surveyed the tracks of the Great Western Railway, toward Bristol and beyond. But Kingdom Brunel also designed the SS Great Eastern, the largest ship in the world at the time of her completion in 1858, specifically designed to carry enough coal to steam to Trincomalee, Ceylon (today’s Sri Lanka), and back, a distance of twenty-two thousand miles. He and Simon Cooper would have had something to talk about, even more so given the Great Eastern’s most famous use: the laying of the first transatlantic telegraph cable, whose 2,700-mile length could be coiled in the massive ship’s hull. Cooper would have especially liked the early transmission rates: $10 per word, with a ten-word minimum. Practically speaking, I was on my way to Porthcurno. But I was aware that really I was on the trail of a broader idea about the triumph of technology over space—and for that there was no better patron saint than Kingdom Brunel.
Within a few hours the tracks looked out over the stormy seas where the English Channel met the Atlantic. Britain was beginning to feel like the island it is. With each mile the view out the window became more nautical. I was headed to the end of the gangplank, a spit of land known as the Penwith Peninsula—the westernmost of the pincers that look as if they’re about to snip at the ships entering the Channel. To my American eyes the landscape was ancient, with cragged trees, roads that sat deep in the fields, and stone farm buildings that seemed to be sinking. Penzance was the end of the line. The beach concessions were all closed for the season but the “Prom”—the seaside promenade—was busy with walkers along the sweeping bay. I rented a car at the station, and since it was midafternoon in fall and I had no appointments for the afternoon, I decided not to bother with a map and aimed only for the sun, feeling my way toward Porthcurno. I figured it would be hard to get lost. There was only one way to go. I had reached the end of the land.
Porthcurno nestles at the base of a valley, a few dozen tidy houses huddled along a narrow lane that ended at a spectacular beach, a short crescent beneath high cliffs. The vegetation was nearly tropical, with scrub trees and flowers, and the water was turquoise. The Falmouth, Gibraltar, and Malta Telegraph Company landed its first cable here, to Malta, in 1870. The beach was chosen over Falmouth, forty miles east, after concerns that the cable there could be damaged by anchors in the busy port. (Cooper would have done the same thing.) Within a few years, two hundred thousand words were transmitting by telegram through Porthcurno annually, and new cables were planned. By 1900, Porthcurno was the hub of a global telegraph network that linked India, North and South America, South Africa, and Australia. By 1918, 180 million words were passing through the valley annually. At the start of World War II, Porthcurno—or “PK,” in telegraph notation, a reference to its original name of Porth Kernow—was the largest cable station in the world. The company by then known as Cable & Wireless operated fourteen cables from the valley, totaling 150,000 miles in length. To protect them from Nazi sabotage, flamethrowers were installed on the beach, and miners were brought in to hollow out the granite hillside and move the station underground. After the war, Cable & Wireless took over the expanded facilities as its training college. Employees from all over the world converged in the valley for classes, to learn how to run the equipment and the business, before being posted to Cable & Wireless stations overseas. The school was active as late as 1993, seeding a close fraternity of men who still fondly recall their days in Cornwall. Porthcurno is their spiritual home, and today the bunker houses the Porthcurno Telegraph Museum, where much of the original equipment is on display and history videos play on a loop.
That evening I was one of two diners at the Cable Station Inn, the pub that occupies the training college’s old recreation center, purchased by its proprietors directly from Cable & Wireless. My journey didn’t seem strange to them. Their neighbor—and a good customer, it sounded—managed one of the landing stations and was a bit of a know-it-all. “He’ll talk your ear off, explaining everything, but he knows more than anyone about this sort of thing,” the publican told me. “Google googles him!”
/> “Maybe he can give you a tour?” his wife ventured.
“No, it’s not such an easy thing,” he corrected her.
I visited the archives at the Telegraph Museum the next morning. A pensioner working her way through a stack of Porthcurno’s old school registries let out a shriek: her uncle was born before her grandparents married. I sat at a long wooden table in the old college building as Alan Renton, the archivist, pulled out boxes of documents from early cable landings on the beach and surveyors’ maps of the bay. The engineer’s report of the “Porthcurnow—Gibraltar No. 4 Cable,” lain in 1919, was a testament to competence if ever there was one. The cable ship Stephan left Greenwich with 1416.064 “nauts”—nautical miles—of cable, made by Siemens Brothers, late in November. A few days later, in a gentle northeasterly breeze, she anchored in the cove at “PK” and sent the end of the cable ashore, supported in the water by ninety wooden casks. By 5:20 that evening, the anchor was hoved in, the cable was paying out over the stern, and the Stephan was steaming toward Gibraltar, “all proceeding satisfactorily.” Within two weeks she was in Gibraltar Bay, preparing to land the other end of the cable on a “fine, bright and clear” day. “Completed final tests and advised the managing Director,” the report concluded. Cable laying was already routine (notwithstanding the engineer’s complaints about “the obvious risks of laying cables in deep water in Winter time on crowded seas and the fact that the Stephan is difficult to handle”). It was a reminder that Porthcurno had already been the communications center of a humming empire for two generations then—and would be for a long time to come, if more quietly so.