Facebook has been calling for low-power server chips for years. Have they finally arrived?

Jonathan Heiliger is the kind of guy you want running your data center. He’s been stringing together servers since the late 1990s, when he co-founded the high-end data center company, Frontier GlobalCenter. Until just a few months ago, he was vice president of technical operations at Facebook, building out the social network’s infrastructure as it skyrocketed from 50 million users to 750 million.

In June 2009, Heiliger shared his thoughts on the latest server chips from Intel and AMD at a conference in San Francisco. And what he had to say wasn’t nice. Hardware vendors had failed companies like Facebook, he said. And the performance gains that their latest generations of processors were supposed to deliver? Facebook simply wasn’t seeing them.

He was annoyed with the server-makers too. “You guys don’t get it,” he said. “To build servers for companies like Facebook and Amazon and other people who are operating fairly homogeneous applications, the servers have to be cheap and they have to be super power efficient, and that doesn’t just mean putting in a really highly efficient power supply, but it means going all the way down basically from starting at the wall outlet all the way to the processor and figuring out how to optimize that power path.”

Heiliger did concede that one company was good at building computers: “Google has done a tremendous job of building and designing their own servers,” he said. Not exactly a ringing endorsement for the status quo in the server industry.

Today, Intel dominates traditional enterprise server rooms, and businesses generally buy their servers from the likes of Dell and HP. But after Heiliger’s harsh words for the traditional players and similar noises from others in the know, some think that the big name server and server chip manufacturers will soon come under threat from upstarts that build a very different type of system — one that’s specifically designed for the “cloud” services such as Facebook and Google and Amazon.

One threat may come from server chips based on the ARM architecture — processors much like those used in the iPhone, the iPad, and so many other mobile devices. The key to ARM’s success on smartphones is power — or, more accurately, lack of it. ARM designs low-power chips that work well enough without burning up a lot of juice. As Jonathan Heiliger made so clear, that’s just as important in the Internet data center.

ARMing up for a server war

Cloud service providers operate data centers that work almost like ant farms, with tens of thousands of servers analyzing and moving vast quantities of messy unstructured data from Web pages to computers on the Internet. Using using a new generation of software technologies such as Memcached and Hadoop, these cloud providers have embraced a whole new way of programming, one that’s fundamentally different from the way software has traditionally been written in the enterprise.

With old-school enterprise software development, IT staffers would make the Oracle database run faster by building bigger computers with more chips and faster processors. That doesn’t work on the cloud. Heiliger made Facebook run faster by throwing thousands and thousands of servers into his data center and having each one bear a tiny share of social networking sight’s growing load.

Now, ARM and other new-age hardware makers believe they can take on the server market by building machines that are good enough to run these cloud workloads, but use far less power than a typical Intel or AMD processor.

Research firm Gartner calls these “extreme low energy servers.” Low-power startups such as Tilera and SeaMicro are already hard at work on such systems, and over the next year, more companies will jump into the low-energy market. Some, including SeaMicro, use Intel’s low-power Atom chip design, but many of them will license the ARM architecture that’s proven so successful on mobile devices.

“The pace of change for at least the next five years is expected to be very fast in designs, capabilities, products and vendors of extreme low-energy servers,” wrote Gartner Analyst Carl Claunch in a research note he released last week.

A new chip for HP’s Mooonshot

The most high-profile of this new round of ARM-baseed companies is Calxeda, a three-year-old Austin, Texas, startup that is building chips that will use much less energy than Intel’s lowest-power Atom processor. On Tuesday, Calxeda introduced its first chip, the EnergyCore. Expected to be ready for server-makers in mid 2012, it will use just 1.5 Watts of power. Compare that to the Intel’s N570 Atom processor, used by low-energy chipmaker SeaMicro. It burns 8.5 Watts. Using a Xeon processor bumps things up to 45 Watts.

The Calxeda chips aren’t designed to be blazing fast, but they should be fast enough for the cloud’s ant farms. And they’ll run so coolly that server-makers will be able to cram a lot more of them into each system. They’ll also have a lot more memory built in than the ARM processor in your iPhone and will ship with built-in networking and management features so they can do much more work, quickly, before having to send or receive data from another part of the server.

That’s where performance bottlenecks can happen, says Karl Freund, Calxeda’s vice president of marketing. “They’re not designed for maximum performance,” he says. “The goal has always been price-performance.

Calxeda isn’t saying what its chips will cost, but the company has secured some big-league backing. Hewlett Packard says it’s designed a 288-chip Calxeda rack-mounted server that’s just seven inches high.

The Intel Downplay

Intel believes that these extreme low energy systems represent only about six to ten percent of the server market, but this kind of cloud-based computing is growing really fast — too fast to be ignored.

That’s why HP is getting into the game. On Tuesday, it announced plans to open a new extreme low-energy server lab in Houston next year, called the Discovery Lab. Part of a program called Project Moonshot, HP wants to give customers a place to test out their software on these new low-power systems before they come to market.

That testing will be important because ARM-based servers can’t run the same popular server programs that are already widely used in the enterprise without some tweaking. The software needs to be recompiled for ARM, and that means that at first only a small number of highly customized programs are going to run on these servers.

Red Hat and Microsoft haven’t said much about how they’re going to support ARM servers. Right now, Canonical’s Ubuntu Linux is the operating system of choice for extreme low-energy systems using chips such as Calxeda’s. Canonical — best known for its desktop version of Linux — recently started development on a version of Ubuntu for ARM servers.

With so many new moving parts — new chips, new compilers, new versions of Linux — these emerging ARM systems will have a lot of potential points of failure. That’s going to be a turn-off for most enterprise IT developers, says Jason Hoffman, chief technology officer with Intel-funded cloud service provider Joyent.

But if someone could build a highly integrated server appliance out of these chips, they might have a winner. “If someone took a Samsung- or an Apple-like approach to making an ARM server — a highly integrated, very beautiful device that has real benefits over the clunky Nokia phone (type of device) that you had before. If someone can do that, it’s very interesting.”