Jason Waxman of Intel (left) and SeaMicro CEO Andrew Feldman show off the new SM-10000-XE server, the first SeaMicro server to use Intel Xeon chips.

SeaMicro has adapted its many-core server design to work with Intel Xeon processors, significantly expanding the type of workloads that can run on the company’s low-energy servers.

The new SM-10000-XE servers use 64 quad-core 2.4 GHz Xeon processors and low-power Green DDR3 memory from Samsung, bringing them together in a 10U server. SeaMicro says the new server will use half the power and a third of the space of equivalent computing power in competing rackmount units.

“Today we have announced the lowest-power, highest-density, highest-bandwidth Intel Xeon–based server ever built,” said Andrew Feldman, CEO of SeaMicro. “SeaMicro now brings the benefits of micro servers – efficiency and massive density – to small and larger-core workloads and to all parts of the scale out data center. Combining the SM10000 architecture with the Samsung Green DDR3 memory and Intel Xeon processors, SeaMicro now sets a new bar for energy efficient compute in the data center.”

Many Cores, Working Together

SeaMicro’s first-generation servers used a mesh of up to 768 Intel Atom processors, which were initially deployed in mobile devices and use far less energy than traditional server chips. By harnessing a large number of low-power processors to work together, SeaMicro’s SM-10000 offered dramatic reductions in power and space usage for that level of computing power. The use of the low-power “wimpy” cores made it ideal for Web traffic and analytics, but not appropriate from other workloads.

Adapting the SeaMicro architecture to use “brawny” Xeon chips expands its use to applications using Java and PHP, memcached caching and MySQL databases. “We want to bring these benefits to all parts of the scale-out data center,” said Feldman.

Mozilla an Early Adopter

One example: Mozilla Corp., the developer of the Firefox web browser, which is using two of SeaMicro’s new Xeon servers in its production data center in Santa Clara, Calif. and will soon install another in its Phoenix facility.

“By offering microservers with small cores based on Atom and large cores based on Xeon, SeaMicro enables us to match our workload to processor type across the data center,” said Matthew Zeier, director of IT Infrastructure and Operations at Mozilla. “With these two types of processors in the same system architecture, SeaMicro is able to meet the computational needs of our entire data center while driving down operating expenses by saving us power and space. We are a very happy customer.”

SeaMicro uses i/o virtualization technology that allows it to remove many of the components seen on a traditional server. I/O virtualization allows companies to save money by using fewer cables and network interface cards (NICs) to connect to networks and storage. Samsung’s Green DDR memory is smaller and uses less energy than competing memory, gelling nicely with SeaMicro’s focus on a sleeker architecture.

The Xeon cores are tied together through SeaMicro’s updated Freedom Supercompute Fabric, which consists of multiple ASICs working together to delivering a low latency, massive bandwidth fabric with low power draw.

“No Segment Untouched”

“The Freedom Supercompute Fabric is the only fabric that can support large and small workloads, enabling SeaMicro to plot a trajectory of servers that leaves no segment of the data center untouched,”  said Zeus Kerravala, senior vice president and distinguished research fellow at the Yankee Group. ”Their rate of innovation is simply dazzling.”

Feldman said he expected the new servers to be SeaMicro’s most popular product. “We will definitely sell more Xeons ervers (than Atom servers),” he said. “Atom is newer as a footprint, so it’s slower to be adopted. We’ve now adapted the same Xeon core people ar ealready using in a smaller footprint.”

SeaMicro says the SM10000-XE are “plug and play” – customers can deploy it without modifications to existing operating systems, application software or management tools. The SeaMicro SM10000-XE is now generally available in the U.S. and at select international locations. The U.S. list price for a base configuration is $138,000.

This has created some new opportunities within the supply chain, as noted today by Cade Metz at Wired Enterprise in an article titled Mystery Men Forge Servers for Giants of the Internet, which examines the role played by server customizers like Hyve and ZT Systems.  Metz tagged along with Joyent CTO Jason Hoffman on a recent visit to a Hyve facility in Fremont, Calif. Read the full article at Wired Enterprise.

Today Emerson Network Power has put forth an answer to this question in “State of the Data Center 2011,” a report and infographic that summarizes data points about data growth, data center facilities and the cost of downtime. Emerson says there are 509,147 data centers worldwide, with 285.8 million square feet of space. Or in more familiar terms, there’s enough data center space in the world to fit 5,955 football fields.

What’s the usefulness of this number? Many providers of products and services for the data center sector are keenly interested in the size of the potential market. Emerson Network Power says the data illustrates the critical importance of data centers in the global economy, and the need to keep them online – which is Emerson’s business.

Data Centers as “Unsung Heroes” of Internet Boom

“Over the last several years, advances in technology, an increased reliance on the internet and social media as well as an increased focus on energy management initiatives have had a significant impact on the data center world,” said Scott Barbour, business leader of Emerson Network Power. “Data centers are the unsung heroes. This infographic illustrates how our reliance on them has grown exponentially.”

This year there will be 1.2 trillion gigabytes (GB) of data created, equivalent to 75 billion 16 GB iPods. That’s more than enough for every person on earth to own 10 iPods. All that data has to live somewhere, driving the need for more data center facilities.

The growing dependence on the data center means growing consequences of downtime, Emerson says. A study the company commissioned last year by the Ponemon Institute found that the data centers average 2.5 outages per year, with an average outage duration of 134 minutes. If that trend tracked over the global footprint of more than 500,000 data centers, that works out to 2.84 million hours of annual data center downtime.

What’s the cost of that downtime? Emerson uses a figure of $300,000 an hour as the business cost of an outage, which translates into a total loss of $426 billion a year. That certainly reinforces the value of uptime.

Emerson’s reports provides some fun with numbers. Is the estimate of 509,147 data centers in the ballpark? Or is the actual number of data center unknowable?

You can see the full infographic on the Emerson web site.

The EnergyCard, a four-node configuration of the Calxeda EnergyCore server which was announced today.

After months of anticipation, Calxeda today unveiled low-power server technology adapted from cell phones that will serve as the cornerstone of a new line of servers for HP.

Calxeda’s EnergyCore is based on processor technology from ARM and consumes about 5 watts of power – and as little as 1.5 watts in microserver configurations – which could allow data center operators to dramatically slash their power usage for some applications.

Calxeda, a startup based in Austin, is among a group of next-generation server and chip makers hope to win business from cloud computing providers by slashing their energy bills. Calxeda, along with rivals SeaMicro and Tilera, use an approach that harnesses thousands of low-power cores that work together on computing tasks. Claxeda has gained attention for its focus on ARM chips, which are  used in many mobile devices like the iPhone and iPad.

“A New Breed of Server”

Calxeda’s approach, which it calls a Server-on-a-Chip (SoC), combines a low-power processor, a network fabric and management software. The integrated fabric switch on every chip lowers cost of interconnecting thousands of servers, while an embedded management engine optimizes power management.The company says this holistic approach has created a “completely new breed of server” that captures efficiencies at multiple levels of the data center energy challenge.

“We believe a new era of energy-efficient servers is now dawning for scale-out workloads, and today, we are introducing the foundational architecture that will enable this breakthrough,” said Barry Evans, CEO and co-founder of Calxeda. ”While we are proud to launch our Calxeda EnergyCore processors, we are even more thrilled with the many partners who are joining us on this journey.”

Chief among them is HP, one of the world’s largest server vendors, which is using Calxeda’s technology in its new Redstone server platform, the first phase of a broader focus on low-power server architecture known as Project Moonshot. The HP-designed system contains 288 Calxeda servers in a single 7-inch (4Rack Unit) chassis.

“A single rack of HP’s Calxeda servers delivers the throughput of some 700 traditional servers, and dramatically simplifies the the infrastructure needed to hook them all together and manage the cluster,” Evans said.

Switch Fabric a Key Component

The EnergyCore processor SoC includes ans 80-Gigabit fabric switch, and an integrated management engine with power optimization software on a single piece of silicon. The EnergyCore SoC design will enable Calxeda’s system vendor customers,to offer a complete server node that consumes only 5 watts, including 4GB of ECC memory and a large capacity SSD. Calxeda said two-node microserver configurations had used as little as 1.5 watts of power in internal testing.

Calxeda said its EnergyCore fabric could offer cost savings opportunities by streamlining network architectures. The company said its fabric acts as a distributed 2-layer switch, which could eliminate the need for top-of-rack switch ports and cabling.

Calxeda’s ARM-based chips are 32-bit, but ARM says it will adopt 64-bit capabilities in future products.

Calxeda says it is not trying to compete with x86 servers from Intel and AMD on a per-thread performance basis, instead focusing on helping workloads “scale out” to use parallel processing across many cores to deliver massive throughput. The company says its performance models indicate that its 1.1GHz product will provide 25 to 50 percent of the processing power of a 2.4GHz x86 4-core processor.

Ideal for Analytics, Web Serving

As a result, Calxeda technology is not suited for every application. EnergyCore chips are ideal for workloads such as web serving, Big Data applications, scalable analytics such as Apache Hadoop, media streaming, and mid-tier infrastructure such as caching and in-memory scalable databases. One of Calxeda’s early customers is the financial services firm Cantor Fitzgerald.

“We are evaluating the Calxeda technology in hyperscale throughput computing for data and simulation intensive applications,” said Niall Dalton, Director of High Frequency Trading at Cantor Fitzgerald. “The Calxeda Linux platform enables rapid porting of our software, enabling us to quickly leverage the energy-efficient ARM cores and Calxeda’s scalable communications fabric to scale our applications to new heights. Many companies in our industry are constrained by space and power, yet our appetite for analysis is insatiable. “We need a 10X breakthrough, and this could be it.”

Earlier this year Calxeda announced its Trailblazer Initiative to create an ecosystem of partners to develop solutions for its server technology. Chief among them was Canonical, developer of the Ubuntu Linux distribution.

“The fundamental constraint in the world of massively parallel approaches to data management and analytics is power,” said Mark Shuttleworth, sponsor of the Ubuntu operation system, and founder of Canonical. “Today marks the beginning of a new way of thinking about what is possible in data and analytics. What is happening here at Calxeda and HP is the beginning of that revolution in the data center.”

“All the stars are in alignment: web 2.0 data-driven businesses, cloud computing, open source portable software, power consumption at crisis levels, and the emergence of server-class performance of ARM processors, ” said Evans.

Your eyes aren't playing tricks on you. It's new hardware with both the HP and Cisco logos, co-engineered by the two rivals.

Are Cisco and HP rivals? Or partners? The two IT giants have been fierce competitors of late, but on Friday they showed that they can still work together when their interests align. HP and Cisco announced the Cisco Fabric Extender for HP BladeSystem, also known as the Cisco Nexus B22 Fabric Extender (FEX) for HP.

The new product, co-engineered by HP and Cisco, allows customers to easily connect and configure their HP BladeSystem c-Class infrastructure and Cisco Unified Fabric. Here’s a look at some of the notable commentary on Friday’s announcement:

• Cisco Blog – Sashi Kiran from Cisco: “The Cisco Fabric Extender for HP BladeSystem was not designed to compete with either company’s offerings. Its purpose is to allow our customers the freedom to mix HP server blades with Cisco Networking. Customers are free to evaluate the merits of other HP and Cisco offerings and determine the best products for their needs.”
• Network World – It’s a curious collaboration, considering that Cisco and HP have become very vocal rivals since Cisco’s entry into the data center server market and HP’s acquisition of networking pioneer 3Com. Indeed, it would seem to undermine both companies’ respective efforts in those market expansions. And HP is a leading provider of blade switches for data center servers. “I’m a little surprised,” says Zeus Kerravala, principal analyst at ZK Research. “I thought they would have gone to anyone but Cisco.”
• CRN – Wendy Bahr, senior vice president of Cisco’s Global Strategic Partner Organization, acknowledged that the collaboration with HP is an eyebrow-raiser, especially given the steely competitive rhetoric that’s been going back and forth between HP and Cisco all year long. The truth is, however, that a lot of enterprise customers have both HP and Cisco in their data center environments, Bahr said. It makes sense for those customers and for the solution providers that serve them to preserve those investments, and make money for everyone in the process.

A wafer displaying Tilera chips. The company announced plans to double the number of chip cores on a processor every two years.

Facebook has prototyped new hardware that can dramatically slash the power required to run web applications like Memcached that support some of the Internet’s largest sites, the company said today. The tests found that low-power processors from Tilera could get three times the performance-per-watt of x86 servers when running key-value store applications.

The research, which is being published today at the International Green Computing Conference, is significant on several fronts. It illustrates how Facebook and other huge data center operators are test-driving new hardware to address their toughest workloads and slash their power bills, and are looking beyond the x86 architecture that has traditionally dominated data center computing.

Boost for Tilera Architecture

The study also provides a major boost for Tilera, which uses a low-power, many-core approach. The testing was conducted using the Tilera-based S2Q server built by Quanta Computer. which is also the manufacturer of Facebook’s Open Compute platform. The server incorporates eight of Tilera’s second-generation TILEPro64 processors into a 2U form factor.

“Our experiments show that a tuned version of Memcached on the 64-core Tilera TILEPro64 can yield at least 67% higher throughput than low-power x86 servers at comparable latency,” says the paper, authored by Mateusz Berezecki, Eitan Frachtenberg and Mike Paleczny of Facebook and Tilera’s Kenneth Steele. “When taking power and node integration into account as well, a TILEPro64-based S2Q server with 8 processors handles at least three times as many transactions per second per Watt as the x86-based servers with the same memory footprint.”

Key-Value Stores Key to Big Infrastructure

Key-value stores play an important role in many large websites. The most prominent example is Memcached, which is used by Facebook, Zynga, Twitter, Wikipedia, Flickr, YouTube, Digg, WordPress and Craigslist. Memcached is important enough to Facebook’s operations that CEO Mark Zuckerberg gave a Tech Talk on it in late 2008.

The Tilera-based server and the competing x86-based servers both met Facebook’s latency requirement of processing the transactions in less than one millisecond. Both platforms had comparable latencies, but Tilera’s processors performed more transactions per second at the required latency.

Memcached is used to provide fast response times to users by keeping data in memory rather than on drives. The authors noted that Tilera’s TilePRO64 architecture ideal for Memcached and other key-value stores workloads because it combines the low-power consumption of slower clock speeds with the increased throughput of many independent cores.

Some Customization Required

However, those performance gains required some customization and tweaking of Memcached, including splitting data tables into separate shards to allow more effective parallel processing, a step required to adapt to TilePRO 64′s 32-bit instruction set. While this is within the capabilities of many large companies using Memcached, it’s not necessary with x86-based architectures.

Tilera uses an architecture that eliminates the on-chip bus interconnect, a centralized intersection where information flows between processor cores or between cores and the memory and I/O. Instead, Tilera employs an on-chip mesh network to interconnect cores. Tilera says its architecture provides similar capabilties for its caching system, evenly distributing the cache system load for better scalability.

Facebook plans to run the same study on Tilera’s new 64-bit Gx3000 series, which was announced in June and will begin sampling later this month.

Facebook’s hardware research is a reminder of the growing role of original design manufacturer (ODM) companies like Quanta, which is an investor in Tilera and is emerging as a key player in the market for custom cloud servers. Quanta is also rumored to be a source for custom servers at Google, prompting recent reports that Google is test-driving servers using Tilera chips.

The new SeaMicro SM10000-64HD server packs up to 768 Intel Atom cores into a 10u chassis.

Pushing the boundary of  high-density computing, SeaMicro today introduced a new version of its next-generation server that packs 768 Intel Atom cores into a 10u chassis. The SM10000-64HD (high-density) extends SeaMicro’s mission to develop servers with improved energy efficiency that can pack more computing power into every square foot of data center space.

SeaMicro is one of several closely watched initiatives to adapt low-power processors in servers to help manage soaring power use in data centers. The Sunnyvale, Calif. company makes x86 servers powered by scores of Atom chips, which have been widely used in mobile devices.

The new server is optimized for cloud computing companies deploying armadas of servers to manage huge web traffic. “The rate of growth and type of work being done by computers is being defined by Facebook and similar Internet companies,” said SeaMicro CEO Andrew Feldman.

Boosting Density from 512 Cores to 768 Cores

SeaMicro’s initial server featured 512 Intel Atom cores working together in a networked computing fabric. By harnessing a large number of low-power processors to work together,  the SM-10000 offered dramatic reductions in power and space usage for that level of computing power. The SeaMicro SM10000 is in production at companies like Skype, Mozilla, eHarmony, France Telecom,Rogers Wireless and China Netcom.

With the SM10000-64HD, SeaMicro has boosted the compute density by re-engineering its basic 5-by-11 inch server board to accommodate six Atom  1.66 GHz N570 dual-core processors, instead of the four chipsets seen in its initial server.

The increased density is not without tradeoffs – the new server uses 3.5 kilowatts of power, compared to 2 kilowatts for its predecessor. While the physical footprint remains the same – the SM100000-64HD is heavier. But Feldman says the gains in computing power outpace the additional total power and weight in the 10U chassis.

14 kilowatts per Rack

Customers can pack four SM10000-64HD into a standard rack, providing  3,072 Atom cores in a 14 kilowatt rack. SeaMicro says this configuration allows the SM10000-64HD to replace 60 traditional servers, four top of rack switches, four terminal servers and a load balancer while using one-fourth the power and taking one-sixth space — all without requiring any changes to software.

“SeaMicro continues to drive innovation in the microserver market,” said Jason Waxman, general manager of Intel’s Data Center Group. “Intel’s Atom dual-core processor provides 64-bit OS support and four gigabytes of addressable memory all while maintaining the benefits of x86 software compatibility. This processor, combined with SeaMicro’s unique server architecture, enables a solution with more than 3,000 cores and 5,100 GHz in a standard rack.”

More Frequent Technology Updates

As it pushes for higher-density and lower power usage, SeaMicro also aims to accelerate the pace of server innovation by providing regular updates to its technology.

“SeaMicro has delivered three server systems in nine months,” said Zeus Kerravala, senior vice president and distinguished research fellow, Yankee Group. “Over that time, the company has increased density by 150 percent and increased its lead in the all important compute per-watt metric. This rapid rate of innovation is a testament to the flexibility and extendability of its system architecture particularly its supercompute style fabric and system software.”

Evaluating Other Processors

The SeaMicro architecture can support any type of processor. Feldman says the company found Atom to be the best starting point, but is evaluating other processors.

“In our lab, we have under development nearly every imaginable CPU that copuld be used in a server,” said Feldman. ‘Bigger CPUs, smaller CPUs, you name it. We are constantly talking to customers about their needs, as can adapt as we need. That’s the beauty of the architecture.”

The SeaMicro SM10000-64HD is available now in the U.S. and at “select international locations.” The US list price for a base configuration is $237,000.

The board for the SeaMicro SM10000-64HD, with six Intel Atom Chipsets on the bottom hald of the board and SeaMicro's customer ASICs at the top.

Calxeda CEO Barry Evans provides the first public display of the ARM servers the company is developing.

Is the growing influence of huge cloud data centers driving a “sea change” in server design and processor technology? Or will x86 offerings from Intel and AMD continue to dominate, despite the current excitement about new architectures and custom server design?

Four companies debated the future of the server and processor market last week at Structure 2011, including executives from AMD and three newer players in the server game – SeaMicro, Tilera and Calxeda. All agreed that huge purchases by cloud computing providers are changing the server market.

“Our projections indicate that while the server market as a whole will grow, a significant part of that will come from cloud,” said Don Newell, the Chief Technology Officer, Servers at chipmaker AMD. “The technology and platform requirements differ pretty dramatically (from enterprise customers). Part of our task is to understand the breadth of that.”

Mega-Data Centers Matter
“The mega-data center guys are becoming so important so quickly,” said Andrew Feldman, CEO of Sea Micro, which makes many-core servers using the Intel Atom processor. “These people are driving the market. Their data centers may not look like yours. But in four or five years, that’s the way you will be building your data centers.”

For the largest data center operators, the status quo has not been good enough. Both Google and Facebook are building their own custom servers, working with original design manufacturers (ODMs) rather than buying direct from mainstream server vendors.

That’s the target market for Tilera, which just announced a new generation of chips with up to 100 cores. Tilera has focused on these huge cloud customers, and envisions that many of its customer servers will be built by ODMs like Quanta Computer, which built Facebook’s servers.

“That’s one of the main ways forward,” said Tilera CEO Omid Tahernia. “It’s a sign that a change is in place is when the supply-chain shifts.

“We’re currently at a point where legacy is not a good thing,” Tahernia added. “The disruption is here. We are at a point where the dam is going to break. It’s not going to fall in one fell swoop. But 5 to 10 years from now, will have a very different conversation.”

Lower Energy Use a Selling Point
Next-generation server and chip makers hope to win the hearts and minds of cloud providers by slashing their energy bills. Tilera, Sea Micro and Calxeda all promise substantial power savings over current offerings from Intel and AMD.

“What we see today is that people have a big problem: they have no space and power available,” said Barry Evans, CEO of Calxeda, which is developing servers based on the low-power ARM chips, which like Atom are used in many mobile devices. “What we hear is that they have budget and want to buy servers, but they don’t have space or power. The current technology just can’t take it. This notion of ‘let’s keep doing it the way we’re doing it’ doesn’t work anymore.”

Show and Tell at Structure
Calxeda has yet to launch production hardware, but at Structure Evans showed off a small motherboard for a Calxeda server. SeaMicro’s Feldman also displayed one of his company’s motherboards, prompting Evans to note that it was bigger than the Calxeda board.

“It’s bigger, but it’s here, and it’s available and you can buy it now,” said Feldman.

SeaMicro’s server architecture combines low-power CPUs, compact motherboards and an interconnection and switching fabric. This approach allows it to pack 512 Atom CPUs into a 10U form factor chassis. SeaMicro’s servers can use any type of chip, Feldman says, but its research  found Atom to be the right place to start.

“What we found is that the Atom architecture is best equipped for Internet workloads,” said Feldman. “It was x86, which allowed us to benefit from existing infrastructure. Intel made some modifications for us and we’re now running 64-bit atom chips. We don’t have a religious view. Our customers are asking for x86.”

That raises the prospect that someday several of the next-generation players might work together.

“We see SeaMicro as a potential customer,” said Tilera’s Tahernia.

Skepticism from AMD
AMD’s Newell acknowledged the market interest in his new rivals, but compared it to previous enthusiasms about Java offload engines. “There were people running around saying ‘there is a sea change,’ ” said Newell. “I’m skeptical that’s going to be anything more than nibbling around the edges. The argument for smaller, less powerful cores depends upon having a larger number of processors to deliver the same performance.”

Newell said that AMD customers can expect a “substantially more aggressive push to take advantage of the level of scale these cloud providers are building out. We’re certainly looking at that and trying to understand how we make servers cheaper and data centers more efficient.”

He also asserted that AMD can adapt if the market shifts to the new many-core competitors. “There’s nothing magic about the architecture that we can’t do,” he said.

“Then why didn’t you do it?” Feldman asked. Earlier in the panel, Feldman said that “the server guys haven’t put in any value for a long time. SeaMicro is a reaction to that.”

Newell said the talk of a server sea change was premature. “It’s interesting to hear people talk about x86 as though it’s a thing of the past,” he said. “Anyone who thinks x86 will not dominate is confused.”

The specs for the new family of Gx 3000 processors from Tilera (Click for larger image).

Tilera today is unveiling a new line of many-core processors optimized to help the Internet’s largest players slash their power bills. The TILE Gx-3000 chips are optimized for companies that are customizing their data center infrastructure to deploy applications across thousands of servers – players like Google, Facebook and Zynga.

Tilera is among a closely-watched group of companies providing low-power chips and servers targeting some of the biggest challenges facing Web-scale companies with mounting power bills. Tilera says its new 64-bit processors, which pack up to 100 cores into a single node, offer better power efficiency than current x86 chips and more computing muscle than new next-generation rivals based on Atom and ARM chips.

The company clearly believes that in offering an alternative to Intel and AMD x86 chips, the largest data centers are the best place to  establish a beachhead.

“We’ve been working with top data centers to optimize for their requirements,” said Ihab Bishara, the director of marketing at Tilera. “We think it’s these top guys who are willing to bring in a new architecture. They are doing a lot of customization at the board level with processors. We’re working with their supply chain.”

Partnering With Quanta

That includes Quanta Computer of Taiwan, which is an investor in Tilera and is emerging as a key player in the market for custom cloud servers. Quanta is an original design manufacturer (ODM) that customizes notebook and server designs for large customers. These include Facebook, which chose Quanta to build thousands of custom servers for its new Oregon data center. Quanta is also rumored to be a source for custom servers at Google, prompting recent reports that Google is test-driving servers using Tilera chips.

Without naming names, Tilera says the TILE-Gx 3000 family was “co-developed with some of the world’s largest Internet brands” over the past two years. Tilera says each Gx 3000 core consumes less than 0.5 watts of power. Bishara said a full rack of Tilera-based servers would require about 8 kilowatts of power, while offering significant improvements in performance-per-watt.

“The Gx-3000 series hits the sweet spot of high performance and low power,” says Troy Bailey, VP of Marketing at Tilera. “Our cloud customers are very excited about the Gx-3000 series. Many have already placed orders.”

Solving Architectural Traffic Jams

Tilera has been shipping many-core processors since 2007. Tilera uses an architecture that eliminates the on-chip bus interconnect, a centralized intersection where information flows between processor cores or between cores and the memory and I/O. Instead, Tilera employs an on-chip mesh network to interconnect cores. Tilera says its architecture provides similar capabilties for its caching system, evenly distributing the cache system load for better scalability.

The TILEGx-3000 series includes three processors to address different market segments. The 36-core-based systems will replace single-socket servers; 64-core-based systems will replace dual-socket servers; and the 100-core systems will replace up to quad-socket servers. The new chips will be available for sampling next month, the company said.

Bishara says Tilera’s focus on standardization and support for Linux will appeal to some of these large customers.

“One of the beautiful things about this market is that the software stack is so simple,” he said. “Any new player just needs to be sure they run these applications well. Most of these guys are using a CentOS package, and some maintain their own custom Linux distribution.It will look and feel like any other Linux box in their data center. “

Calxeda uses  chips optimized for cell phones to make low-power servers that can slash the energy costs for data center operators. As energy efficiency becomes a leading priority in the data center, end users are keenly interested in next-generation server technologies focused on low-power chips such as Intel’s Atom and the ARM technology embraced by Calxeda.

Building an ARM Server Ecosystem
With its Trailblazer Initiative announced today, Calxeda hopes to create an ecosystem that will have ready-made solutions available for its product launch. Calxeda is providing Trailblazer partners and end users with early access to hardware and technical expertise. Calxeda said the program is a response to demand from customers and partners to get its chips into servers and data centers for proof of concept testing.

One key partner is Canonical, developer of the Ubuntu Linux distribution. “We are proud that Ubuntu has been chosen as the operating system to launch the Calxeda Trailblazer initiative,” said Canonical CEO Jane Silber. “We have been working with Calxeda and have been developing Ubuntu Server to run on ARM servers for some time now . This program further aligns our organisations and will help companies and projects looking to build the hyper-efficient datacentres of tomorrow.”

“We see Trailblazer as a critical step in the evolution of this marketplace, and are thrilled by the support we are receiving from the community to help it materialize,” said Barry Evans, Calxeda’s CEO. “The data center energy crisis is real, and the faster we can help our system vendors and partners get complete solutions in our customers’ hands, the faster we can help them save money, and reduce their energy consumption and carbon footprint.”

Other initial participants in the Trailblazer Initiative include:

• Autonomic Resources, a cloud provider focused on solutions for government
• Caringo, which makes cloud storage software
• Couchbase, the NoSQL Database Company
• Datastax, provider of Enterprise Cassandra and Hadoop support solutions and products
• Eucalyptus Systems, with its on-premise cloud computing software platform
• Gluster, which makes open source storage software
• Momentum SI, an IT transformation services specialist
• Opscode, a leader in cloud infrastructure automation
• Pervasive, with its Datarush big data preparation and analytics accelerator

Calxeda is backed by a syndicate including venture capital firms and semiconductor firms, including ARM, Advanced Technology Investment Company (ATIC), Battery Ventures, Flybridge Capital Partners and Highland Capital Partners.

At the recent Structure Big Data conference, we spoke with Karl Freund, the Vice President of Marketing for Calxeda, who provided an overview of the company’s products and roadmap. This video runs about 7 minutes, 30 seconds.