When Eric Frenkiel started an engineering job at Facebook (FB) in 2010, he encountered one of the world’s biggest databases: a stash of hundreds of millions of users’ likes, messages, and photos. After just a year, Frenkiel realized that he could use his experience to build a better data storage system for companies less tech savvy than his employer. He and Nikita Shamgunov, another Facebook engineer, left the social network in February 2011 to create—bravado alert—“the world’s fastest database,” according to Frenkiel, 26. “People thought we were crazy because we left before Facebook went public,” he says. “There we were, just two guys and a dog.”
The engineers set up shop in a San Francisco loft where they lived and worked. “Nikita, God bless his soul, would sleep next to our servers,” Frenkiel says. “He literally slept next to his code.” They named the company MemSQL, in reference to a popular database language, and hired additional engineers to complete a mad dash to debut their first product. In June, the company started selling software that runs about 30 times faster than traditional databases made by Oracle (ORCL), IBM (IBM), and others, according to Frenkiel.
The MemSQL crew had reason to hurry. Their database software needs to stand out among the offerings of dozens of competitors that have appeared in the last five years. Computer scientists have been perfecting data-center technologies that help the biggest Web services—think Google (GOOG)—serve millions of users. The innovations include not just database software but servers, storage systems, switches, and security hardware. Now they’re making their way into non-Web businesses, many of which are struggling to process huge amounts of data or serve smartphone-toting customers and employees who demand on-the-go access.
This burst of innovation has led to something of a golden age for infrastructure technology. A horde of enterprise-focused startups, fueled by surging demand, could go public this year in a trend that trumps the dot-com buildout in scale. “It’s going to be about 10 times larger than what the Internet did,” said Ted Schlein, a managing partner at the venture capital firm Kleiner Perkins Caufield & Byers, during a recent Bloomberg event held in New York. “Just the fact that we have that many more people online with so many more devices…. That’s what is getting everyone excited.” That, and the huge amount of money that could change hands. The market for data-center hardware totaled $100 billion in 2011, according to the research firm Gartner (IT), which predicts that number will surpass $120 billion by 2015. Software represents an enormous sales opportunity as well: Databases alone brought in about $24 billion in 2011, says Gartner.
During the dot-com boom of the late 1990s, the undisputed kingpins in these markets were Cisco Systems (CSCO), EMC (EMC), Oracle, and Sun Microsystems. They sold top-of-the-line gear first to Web pioneers such as EBay (EBAY) and Yahoo! (YHOO) and then to just about every major company readying itself for the Internet era. But after the bubble burst, many big companies decided they’d overindulged and dramatically cut spending. The pace of innovation slowed. “I think enterprise technology kind of sucked for about 10 years,” said Schlein.
Cut to 2012: Net infrastructure is again being upgraded, this time so companies can do computing in the cloud and provide tools to mobile users. Over the last decade, big Web services like Google and Facebook have learned they don’t have to buy expensive gear from Oracle or EMC. Instead, they shackle together cheap hardware and write software to smooth over any rough spots. That’s changed the economics of crunching large datasets or providing “like” buttons for millions. “I look at those companies rhetorically as the Fords and GEs of the Information Age,” says Jim Whitehurst, the chief executive officer of Red Hat (RHT), a major infrastructure software maker.
The modern-day Fords have made much of their work open-source, so that other technologists can use it for free—and that’s restarted data-center innovation. Some of the most promising new enterprise-tech companies borrow techniques from Google or Facebook, tweak them, and sell gear or software to non-Web businesses. Cloudera, a hot enterprise company, makes data analytics software that helps the oil and gas industry, for example, uncover insights from large sets of information. It relies in large part on Hadoop, an open-source technology created by Google and Yahoo.
The enterprise startups are just at the beginning of an IPO wave similar to the one that washed across the consumer Web sector over the past year, when Zynga (ZNGA), LinkedIn (LNKD), and Facebook all went public. Business-computing startups including Splunk (SPLK) and Infoblox (BLOX) have already made their stock market debuts this year. ServiceNow, a modern take on managing data-center operations, is firing up its IPO roadshow machine. Next will be Palo Alto Networks, which makes a smarter firewall capable of coping with security holes introduced by Web services such as Facebook. There’s a handful of other infrastructure players set to follow, as these business-computing stars attempt to give their consumer Web counterparts a run for the money. “That’s what we’re loving,” Schlein says