This is an interesting turn of events.

Lately, startups like Calxeda and SeaMicro have been making waves by enticing data center operators with the promise of potent yet power-sipping servers using low-watt processors that are decidedly non-standard on servers. Namely, ARM- and Intel Atom-based chips, respectively.

Now it turns out that SeaMicro sees some value, or at least expanded market opportunities, in that old server stand-by, the Xeon. Hence the SM10000-XE, a server containing 64, quad-core Sandy Bridge-based Xeons at 2.4 GHz.

The point, according to this HPCwire piece, is to make quick work of mainstream enterprise workloads.

With the addition of low-wattage but more performant Xeons into the mix, SeaMicro is looking to expand the microserver business into what it calls “brawny applications.” That includes more traditional enterprise workloads like Java, PHP, MemCacheD, and NoSQL, as well as web-based database processing. SeaMicro CEO Andrew Feldman characterized the new Xeon-powered SM10000-XE as the “the mainstreaming of the microserver.”

And keeping with the company’s energy- and space-saving roots, the server takes up a third of the space and uses half the power — 3.5 KW on average — of a comparable system comprised of volume rack servers, according to SeaMicro.

All good stuff, but what does it mean for the company’s Atom-based hardware? While SeaMicro’s CEO, Andrew Feldman, expects bigger sales from its Xeon hardware, the tech that put it on the map is doing alright…

Even without the SM10000-XE though, the company has been doing “phenomenal,” according to Feldman.  Although he didn’t offer how much revenue his company collected during their first year of business (2011), Feldman says it was more than the combined sales of Riverbed, 3PAR, Aruba Networks and Data Domain combined during their first year. “It looks pretty bright out there right now,” he says.

A little vague, but good to know.

HP got the industry buzzing this week when it announced plans to build server hardware based on low-power ARM processors. And while it will be a while before Project Moonshot’s tech hits data centers, that’s not stopping folks from dreaming up some big data implications.

Better yet, some enterprising souls are already demonstrating what its ‘cloudy’ future could look like.

Over at CNXSoft,  Noritsuna Imamur of the Open Embedded Software Foundation, shows off Google App Engine running on a Pandaboard cloud cluster.

Norisuna had a comparison table that showed the Pandaboard cloud (OMAP4 @ 1GHz – 12 Cores) uses 4x times [less] space and 5x less power (50W vs. 257 W) than a Xeon 1.8GHz Quad Core server with about the same amount of processing power and memory (6GB vs. 8GB).

Neat!

Check out the video below to see it in action. Favorite part: When the camera pans over to the Kill-a-Watt.

Remember that report about Calxeda and HP teaming up for low-power ARM servers? Well, now it’s official and it even has a name: Project Moonshot.

To get the industry used to the idea of computing on servers built around ARM processors, HP is laying the groundwork with a couple of initiatives. First is Project Pathfinder, which brings together partners including the aforementioned Calxeda along with ARM and Red Hat, to develop the technologies and hardware platforms and drum up industry support. Interestingly, the inclusion of AMD implies that Moonshot will also encompass low-power x86 processors. Hmm…

Also on deck is the HP Discovery Lab, pictured above with its Head of Platform Research & Development, Gerald Kleyn. Set to open near Houston, Texas early next year, customers can use the facility, either remotely or on-site, to test their apps on the The Redstone Server Development Platform. This hardware, ”incorporates more than 2,800 servers in a single rack, reducing cabling, switching and use of peripheral devices to reduce complexity by up to 97 percent,” according to HP.

The potential benefits? A whole lot ‘less.’

By the company’s estimates, IT shops can expect these servers to consume up to 89% less energy, occupy 94% less space and cost 63% less overall. Pretty tempting, but we’ll get to see if those savings bear out when data center operators are able to put the tech through its paces for themselves next year.

For a behind-the-scenes look, watch the HP Discovery Lab video here. (Eagle-eyed viewers will notice the Calxeda logo on the processors when Gerald Kleyn tilts the server chassis toward the camera at about 1m 16sec.) You can also check out HP’s Project Moonshot announcement video below.

Screengrab credit: HP

[via Hacker News]

Things are getting exciting on the computer server front.

For a couple of years, an Austin, Texas-based startup called Calxeda (formerly Smooth-Stone) has been making the case for energy efficient servers using ARM processors — that currently dominate the smartphone and tablet markets — as it prepared to commercialize its tech. Now it looks like the company has found a powerful ally in getting its hardware into data centers.

According to a Wall Street Journal report, HP and Calxeda are partnering to offer ARM-based servers to data center operators and cloud providers — oftentimes one and the same — where low power use and compact physical footprints are emerging as big selling points.

Calxeda isn’t alone in the ARM server race. ZT Systems and Marvell have also set their sights on a market dominated by x86 processor technology from Intel and AMD. But HP’s support, even at this early stage, is a form of validation for Calxeda’s strategy.

It’s also some of the strongest evidence yet that the market is finally ready for servers that won’t put data center operators at a power consumption disadvantage as they scale up to meet massive “cloudy” workloads. According to the WSJ report, sources claim that “compared to mid-range Intel server processors, Calxeda’s chips consume about 90% less energy, take up about 90% less space and have a lower overall cost of ownership.” That’s a pretty compelling selling point.

Lending some substance to these “leaks” are cryptic tweets from Calxeda’s Twitter stream hinting at “Big news from @Calxeda coming soon.” (source) and “Change is in the air. And we don’t mean fall colors.” (source)

The companies are expected to unveil a prototype soon, but don’t expect waves of HP/Calxeda hardware in the immediate future. Volume production isn’t scheduled until the second half of 2012.

Over the past couple of years, some startups have been using mobile chips as a way to bring high performance computing to the data center in the form of low-power, space-saving servers.

On Friday, August 19, one of those startups, SeaMicro, will be discussing the tech behind its Intel Atom-powered SM10000-64HD server (pictured above) during the company’s presentation, “Building Data Center Servers Using Cell Phone Chips,” at the Hot Chips conference at Stanford University. I could get picky and mention that Atom-powered devices are what you could generously call “few and far between in the cell phone/smartphone market” — ARM rules this space — but, well, I kinda just did…

In any case, the SM10000-64HD’s tech is worth discussing, considering that it produces the computational output of a traditional rack of servers using just one fourth of the power in one sixth of the space, according to the company. Using this “rack in a box” architecture, SeaMicro packs 384 Intel Atom N570 dual-core processors rated at 1.66 GHz into an x86-friendly system ($237,000 base price) that the company hopes will entice data center operators and cloud providers.

Why should they care? Because they’ll soon be facing dwindling floor space, exploding energy costs and environmental scrutiny, if they aren’t already. SeaMicro’s founder and CTO, Gary Lauterbach, Founder and Chief Technology Officer, hopes those factors — not to mention his company’s tech — resonate with the crowd at Hot Chips.

He states, “As data centers become more and more essential in day-to-day life, power consumption and space requirements are issues with real economic and environmental consequences. We are excited to share our unique approach to the problems faced by the data center at Hot Chips, in order to spark further discussion and innovation.”

Let’s hope.

Image credit: SeaMicro

Literally, in fact.

Today, Facebook took a dramatically different course than its peers and let the world look behind the curtain. Whereas other big Internet firms consider the “nuts and bolts” details of their IT infrastructure a competitive advantage, Facebook open sourced its custom, energy efficient server and data center tech under the Open Compute Project.

Prineville data center redeemed?

Despite courting controversy for siting its first ever data center in Prineville, Oregon — it’s primarily powered by coal — Facebook has been able to realize huge energy savings. In fact, the Prineville center uses “38 percent less energy to do the same work as Facebook’s existing facilities, while costing 24 percent less.”

This is achieved by implementing free-cooling, Ethernet-powered LED lighting (!) and dumping traditional UPS (uninterruptible power supply) setups in favor of battery cabinets that are sandwiched between “triplet” server racks. Each triplet is comprised of three 42U columns that house 90 servers total and six rack switches. Also in the mix are custom motherboards and 94.5 percent efficient (90-plus technically speaking) power supplies.

All told, the innovations are helping Facebook deliver 93 percent of the power its Prineville facility takes from the grid to its servers and hit and impressively low 1.07 PUE mark — better than the company’s already-enviable 1.5 PUE rating. It also bodes extremely well for the company’s second data center build in North Carolina.

Sure, the specter of coal still hangs over the facility. But by its actions today, Facebook is providing the IT community actionable insights and technical know-how that has the potential for tremendous energy efficiency gains in a data center market that’s growing rapidly thanks to the rising popularity of Web services and clouds. Though many eco-minded geeks would like to see the company embrace renewable energy for low-emissions computing, this is a step in right direction and at the right time. Plus, all this could help add momentum to the clean energy cause by providing firms with a roadmap for future facilities that make the best use of any and all energy sources.

Follow the leader

If nothing else, Facebook has made itself synonymous with efficient IT operations (1.07 PUE!). Even if server vendors and IT managers turn their noses up at the Open Compute Project’s designs and methodologies, Facebook’s influence looms large. Granted, its popularity and market presence may be tough to achieve, but here’s one aspect of its success that any forward-thinking IT department can absolutely replicate and at lower cost than traditional approaches: its data centers.

In short, Facebook is mainstreaming Green IT. What’s not to like?

Read up on the Open Compute Project here. Or check out this video which touches on some of the highlights and explains Facebook’s philosophy on open sourcing its data center tech.

Image Credit: Open Compute Project

Some interesting insights from Facebook this week as the company broadcasts some of the thinking behind its IT strategy. And instead of submitting to the virtualization craze that’s spreading through corporate data centers, the social networking giant is throwing its weight behind microservers.

This comes as Intel announces plans for Xeon- and Atom-based processors for microservers; small, low-wattage servers that are little more than a processor, some RAM, networking interconnects and a smattering of storage. And while they can’t hold a candle to enterprise-class iron performance-wise, they have more than enough horsepower to handle many dedicated and/or low-level tasks.

Tasks like Web serving, according to Computerworld…

Front-end Web servers are where Facebook is nearly ready to start using microservers, according to Coglitore [Gio Coglitore, director of Facebook labs]. “With Intel’s announcement, it’s just about to happen,” he said. Facebook will probably start implementing microservers on a large scale beginning late this year or early next year.

Makes a ton of sense for a company that’s scaling as quickly as Facebook. It also explains why virtualization doesn’t suit the company, as least not at present. When so much hinges on the availability of its systems, it simply can’t afford to risk (or even deal with) losing even a handful of virtual machines if just one server gives up the ghost.

For Intel, microserver architecture is a defense against ARM, which is charting a course toward server chips and currently dominates in the low-power smartphone and tablet processor markets. It’s also a way for the leading chipmaker to get in on the green/cloud server buzz that’s surrounding innovative startups like Calxeda.

And considering how quickly Facebook, Google and others have been growing their IT infrastructures lately — not to mention showing a willingness to explore alternative server architectures — it’s wise to cater to this market. If nothing else, they’re showing no signs of slowing down, which translates into hefty hardware sales.

How many of your servers have been around for a decade? Not many, it’s safe to assume. (None ideally.) Chances are that the servers you deployed in 2001 were replaced 3-5 years later. And if there are a couple stragglers still around, they’re costing you a bundle in electricity and rack space.

Fortunately, longevity is creeping into the server landscape thanks to virtualization and hardware vendors that are all to0 happy to cater to the efficiency-boosting software technology. In his latest post, InfoWorld’s Paul Venezia offers some great insights that are sure to please IT execs. Sure, it flies in the face of folks that anticipate accelerated server replacement schedules because of virtualization, but there’s good reason to hope Paul’s vision plays out.

What lifted the current generation was a combination of mature hardware support for virtualization and the multicore NUMA boosts from AMD and then Intel. Those elements worked hand in glove to get us to a position where most medium-size to large enterprises find they need a remarkably small number of physical servers to run their entire data center. Unless we see a massive rise in software resource requirements, the same state of affairs could hold true for the next decade.

Neat, right? Complicit in this conspiracy to slim down the data center are Intel and AMD (soon ARM?). Both chip companies have made huge strides in crafting virtualization-friendly processors and chipsets, furthering technologies like NUMA (Non-Uniform Memory Access) that make juggling multiple server workloads possible.

This is also proving key in numerous data center consolidation projects, like the one announced today between NYC and IBM. Nestled in the press release were clues about the virtualization-heavy, cloud-ready infrastructure that will replace the city’s creaky, multi-site IT platform. Although not explicitly stated, it’s a given that virtualization — and IBM server hardware that supports it (virtually all of it, nowadays) — will feature heavily and enable the city to reduce 50 data centers to a handful, if not less. If you’re thinking big savings, you’re right.

And if the tech runs well and endures for the next 10 years, that’s great news for cash-strapped NYC — energy savings and reduced carbon emissions aside.

Image credit: Flickr user Tim Dorr – CC (Lost fans, check out those names!)

The ARM-based server buzz continues this week with an announcement from Secaucus, NJ-based ZT Systems.

The server and PC maker unveiled the R1801e, a 1U unit with up to 16 ARM Cortex-A9 processor cores that operates at a mere 80 watts. That includes eight server modules, storage in the form of eight SSDs and dual integrated Gigabit Ethernet switches. According to a company release, each module is comprised of an STMicroelectronics’ SPEAr 1310 microprocessor with dual ARM Cortex-A9 processor cores, which  with the help of PHYTEC America, integrated the SPEAr 1310 with 1 GB of DDR3 ECC DRAM, 1 GB of NAND Flash and Ethernet to form a “System-on-Module” (SOM).

With this move, ZT Systems joins Marvell and Calxeda (formerly Smooth-Stone) in trying to establish a market for servers based on ARM processors. This week, Calxeda not only announced its name-change, it also announced new hires in an effort to bring its tech to market in 2011. Although ARM dominates the mobile market via its low-power mobile chips,  its presence in the server realm remains practically non-existent, despite interest from the IT community. Now that ZT Systems is in the mix, it could light a fire under bigger server vendors. That is, if Calxeda and ZT make a big enough splash and can encourage software developers think beyond x86.

Good-bye Smooth-Stone, hello Calxeda (pronounced “Cal-Zeh-Dah”).

The startup, which specializes in server designs using ARM’s low-power processor technology, is not only undergoing a name change, new staffing developments point to a company that’s gearing up to take on server vendors that have made their fortunes on Intel’s and AMD’s x86 processors. Earlier this summer, the company raised $48 million from VCs and even ARM itself to further the development of its tech.

To help chart a course toward the energy efficient data center of the (near) future, Calxeda attracted the following execs from some pretty big names in IT:

Karl Freund, formerly of IBM and Hewlett-Packard, as Vice President of Marketing
Bob Baughman, previously with Polycom and Marvell, as Vice President of Business Development and Sales
Steve Beatty, Vice President of Manufacturing, formerly of Freescale and SigmaTel.

Calxeda expects to release the first samples of its ARM-based server technology, which the company says will result in a 10X bump in data center performance efficiency, in 2011.