Servers at Capacity

SPEED 010, 6:44 AM. Yoo Jae-won lay in bed with his eyes barely open, still drowsy from sleep. Just days earlier he would have snapped awake around 4 AM, sharp and ready to begin the day. Now it seemed sleep alone no longer fully restored his energy.

“Am I doing this right?” He had launched the project with such confidence, yet a faint doubt crept in whenever his body felt stiff. It was understandable. Since arriving in Redmond, he had discovered that he himself was carrying the heaviest workload. Even the carefully selected Alpha Team could not match his output. The combined work of all fifty members still fell short of what he accomplished alone. When the source code was examined, his portions were superior in every respect—more optimized, more concise, more intuitive, and far more reusable.

He had known from the start that the critical sections would fall to him, but being human, the irritation still surfaced.

“When this happens, the only answer is to get up immediately.” Knowing that lingering in bed would only make rising harder, Yoo Jae-won kicked off the covers and stood. Complaining about what could not be changed only soured his own mood. The developers’ skills improved at a slug’s pace; recharging his own spirits was faster. Nothing restored his mood like Tiffany’s voice, yet she was a heavy sleeper and a call now would be impossible.

“Might as well use the computer.” Computers were both his specialty and his hobby. After pulling on casual clothes, he headed straight for the machine in the living room. The computer was not hotel property. It was not even the portable Shell Book; it was the iWorks he had brought from home. Its solid aluminum housing made it large and heavy, but since he planned to stay until the next Android was finished, installing a full workstation posed no problem.

“Still looks impressive every time.” He had designed it himself, yet the sight never grew old. It was in a completely different league from Apple’s newly released PowerMac. Apple had changed little despite his presence in the market; the PowerMac looked exactly as he remembered. Internally it used an enhanced IBM PowerCPU with HPC technology and carried more memory, but the exterior and MacOS remained unchanged. Market reaction had been lukewarm. Apple retained steady demand in electronic publishing and graphic design, yet failed to expand elsewhere.

“Once this thing launches, it’s over.” The real issue was that year-end release now looked uncertain. He had intended to unveil the iWorks as an intelligent workstation running the next Android packed with new features, but development of that OS was not proceeding smoothly. Releasing with the existing 2.0 version would feel half-finished. Neither he nor TG treated the iWorks launch as an urgent priority, so it was not a crisis. TG was pouring every resource into TG Mobile’s pilot service, determined to crush Shinsedae Telecom from day one and imprint a powerful impression on consumers. Their ‍‍‌‌⁠​​⁠‌​​‍​​‍​‌​⁠‍‌‍⁠⁠‌⁠‌‍‌‍‌‌‌‍⁠​‌​‌​‌‍‌‌‌⁠‍​​‌‍⁠‌‌⁠⁠strategy was to seize first place the moment official service began and hold it. To that end, TG Mobile was expanding coverage with every available means—large relay stations across Seoul and parts of Gyeonggi Province, plus additional repeaters in shadow zones blocked by buildings or mountains. Powerful handsets were also under rapid preparation. The Tiffany Phone Yoo Jae-won had created was already confirmed. TG was simultaneously courting both major electronics conglomerates and technically capable smaller firms to build a portfolio that could satisfy every consumer taste.

The contentious point was the handset supply model. Two options were under discussion: TG Mobile would supply only the network and let consumers buy their own phones, or TG Mobile would purchase handsets wholesale from manufacturers and distribute them through its own dealers. Surprisingly, the debate was not heated. Both electronics firms and telecom operators preferred the second approach. The first gave consumers maximum choice—they could pick any handset and any carrier. The second placed all power with the supplier, much like a gas station also selling cars. In return, manufacturers enjoyed the security of large, immediate orders without inventory risk.

Yoo Jae-won was indifferent to the model. Whether through carrier bundling or direct sales, he intended the Tiffany Phone to dominate as a premium product backed by genuine consumer demand. As TG Mobile’s second-largest shareholder, however, he needed to maximize the company’s profits. Therefore the Tiffany Phone would, for the foreseeable future, be exclusive to TG Mobile in Korea. Shinsedae Telecom appeared to understand this; they had not even requested units for themselves.

“Hmm. So how many units should we prepare?” Production of the Tiffany Phone was already complete. The million-dollar collaboration with the American luxury jeweler Tiffany & Co. was also progressing steadily. Only mass production remained, and the method would depend on TG Mobile’s order volume. Below three hundred thousand units, simple OEM manufacturing by Korean SMEs would suffice. Above that threshold, he was prepared to build ‍‍‌‌⁠​​⁠‌​​‍​​‍​‌​⁠‍‌‍⁠⁠‌⁠‌‍‌‍‌‌‌‍⁠​‌​‌​‌‍‌‌‌⁠‍​​‌‍⁠‌‌⁠⁠a dedicated factory. It would also be wise to establish a small team capable of prototyping additional mobile devices—tablet PCs among them—rather than remaining dependent on TG Computer’s existing lines. One thing was certain: he would never build factories in Taiwan or China. He knew better than anyone that the current rush of companies into China for cheap labor would reverse by the early twenty-first century, when firms would flee in revulsion. Technology and trade secrets transferred to Chinese partners would be handed straight to the authorities, and intellectual property would be valued less than roadside gravel. Chinese factories were already embedding spy chips on boards to siphon corporate and government data without detection.

“They’re already inflicting massive damage.” ID Group was suffering significant losses at this very moment. As China opened, government offices and enterprises were rushing into office automation—meaning computers. Export restrictions had been lifted, and foreign PCs from HP, Compaq, and Dell were flooding in. The real problem lay with Chinese assemblers who imported components and built machines locally. They paid full price for hardware but installed nothing but pirated software. ID Office was the most copied product, followed closely by the Android operating system. Most machines sold by major Chinese firms came preloaded with the adware version; many even bundled cracks to remove the ads entirely. Enterprise editions without adware were simply shipped with outright illegal copies.

“Simple copying would be one thing.” Far worse was the full-package counterfeiting. Thick manuals, CDs, and packaging identical to the originals were circulating widely. Many consumers bought these fakes believing they were genuine.

“Ugh. Maybe I should just drop Chinese support.” Disgusted, Yoo Jae-won briefly considered removing Chinese language support from the next Android. He had no real intention of doing so; the thought was pure irritation. Profiting from software in China was nearly impossible, yet hardware sold at full price remained lucrative. Once China mastered hardware manufacturing, knockoff phones and PCs would proliferate, but the damage would still be less than the software losses. As Chinese incomes rose, demand for legitimate products would grow as well.

After grumbling about China for some time, Yoo Jae-won launched a program the moment iWorks finished booting. It was not ID Talk, his usual first choice, but the early-access version of Ultima Online.

“What is this?” Yoo Jae-won stared in surprise at the message delivered in a solemn voice instead of the usual login screen. It was written in the style of Ultima Online’s world, but the meaning was plain: the server was full and he would have to wait. He checked the time again—7:02 AM, the quietest hour for any online game. Yet a queue existed. Wondering whether maintenance had shut down some servers, he checked the homepage. All servers were listed as operational; every status ‍‍‌‌⁠​​⁠‌​​‍​​‍​‌​⁠‍‌‍⁠⁠‌⁠‌‍‌‍‌‌‌‍⁠​‌​‌​‌‍‌‌‌⁠‍​​‌‍⁠‌‌⁠⁠indicator simply glowed red.

“Jackpot!” The word burst out ‍‍‌‌⁠​​⁠‌​​‍​​‍​‌​⁠‍‌‍⁠⁠‌⁠‌‍‌‍‌‌‌‍⁠​‌​‌​‌‍‌‌‌⁠‍​​‌‍⁠‌‌⁠⁠of him. No empty seats at this hour meant Ultima Online had far more players than he had realized. After approving the large marketing budget for early-access promotion, he had been too busy to monitor results. The game was performing even better than expected. Richard Garriott’s earlier messages about surging applications and breaking ten thousand sign-ups had already told him early access was succeeding, but this level of demand was extraordinary.

“How is Lineage doing?” While preparing Ultima Online’s early access, news had arrived that TJ Soft had signed a contract to use ID Cloud servers. Kim Taek-jun had kept his promise. He had ordered capacity for up to five thousand simultaneous users and one hundred gigabytes of storage—roughly twenty million won per month. Once the servers were ready, TJ Soft began advertising in a style almost identical to Origin Systems’ approach. Game-industry marketing differed little between Korea and the United States, except for one Korean-specific tactic: targeting PC cafes. Before dedicated PC bangs existed, PC cafes had spread rapidly around universities—cafes with public computers where students could also buy coffee. Kim Taek-jun visited these cafes personally, distributing Lineage CDs to build offline awareness.

On May 31, both games launched. Ultima Online opened under the early-access label; Lineage Online launched as an open beta. The key difference was that Ultima Online charged twelve dollars for access while Lineage Online was free.

“Might as well take a look.” The Ultima Online queue was moving far too slowly. Yoo Jae-won launched Lineage Online instead.

“Hmm. This feels comfortable.” In Redmond it was 7 AM; in Korea it was 10 PM, peak hours for computer use. Yet the server felt spacious. Yoo Jae-won nodded. Playing the game himself explained the muted response. It differed in many ways from the Lineage he remembered. Built in 2D as a simple hack-and-slash action RPG, it retained the random stat rolls from character creation, but the overall quality was noticeably lower. Song Jae-kyung, who had held a major stake in the original development, had never joined; only Kim Taek-jun and his amateur team had worked on it.

“Daily traffic usage has also dropped a little.” Although TJ Soft rented cloud capacity, Yoo Jae-won could not arbitrarily inspect or monitor the data—any more than a landlord could enter a tenant’s home uninvited. The only metric available was daily traffic volume, which served as a rough proxy for active users. Usage had peaked at launch and was now trending downward. He did not sneer at the numbers. The game market was full of unpredictable variables. Even the currently dominant Ultima Online could grow stale after official launch, while players might yet flock to the presently quiet Lineage Online. More importantly, both companies succeeding benefited him far more than one side winning outright. Every additional user on ID Cloud meant more shared revenue.

“Hmm. Time to expand the data centers again?” ID Technology currently operated two data centers, in San Francisco and New York. While some countries, including Korea, still ran independent servers on mainframes or small clouds, no one had yet built large-scale facilities. All future internet services from ID Group would run on cloud infrastructure. Warcraft, Quake, and Return to Castle Wolfenstein multiplayer servers from ID Entertainment would all operate inside the cloud. Surplus capacity could also be leased to other internet companies—a major business in its own right. Yoo Jae-won’s ultimate goal was for ID Cloud to host not only ID Group’s services but everyone else’s as well. Any innovative service another company launched would ultimately generate profit shared with him. From a challenger’s perspective, the advantages were obvious: no need to purchase expensive hardware, capital preserved for core development, easy exit if things went wrong, and elastic scaling during sudden success.

“This time, Korea and Europe would be ideal.” A new CPU with expanded instructions would soon launch; it would be perfect for building large cloud facilities in Seoul and Europe.

While he had been thinking, the minimized Ultima Online queue had finally cleared. Yoo Jae-won switched the game to full screen and began logging in. Unfortunately, his playtime did not even reach an hour. Kim Dae-seok had arrived early for work, and they needed to eat breakfast together before heading to the Alpha Team office in Redmond. That pattern—intense concentration all day, late return to the hotel—had already become routine. Thanks to his focus, Android development was advancing at tremendous speed. To the rest of the Alpha Team he looked like a monster. Yet the goals he had set were so ambitious that a great deal of work still remained.

A break in this routine arrived in early June with a single phone call.

“Chairman, you should take this one.” Kim Dae-seok handled every call except those from family, Tiffany, or close friends. In the middle of deep concentration, only truly urgent matters reached Yoo Jae-won. The fact that Kim Dae-seok had not filtered this call meant it was significant. Trusting his secretary, Yoo Jae-won accepted the phone without protest.

“Hello? This is Yoo Jae-won.”

— Good morning, Mr. Yoo Jae-won. This is the International Mathematical Union.

A low, resonant voice came through the receiver. The International Mathematical Union? It seemed the moment he had been waiting for had finally arrived.