SAN FRANCISCO—The much-needed transformation of the data center will go through Intel, which has the innovations, manufacturing capabilities and broad product portfolio that others do not, according to company executives.
The giant chip maker may be trying to make inroads into new markets like mobile devices on the client side, but officials are determined to thwart any significant incursions into the data center by longtime rival Advanced Micro Devices as well as ARM and its lineup of partners, from Samsung to Calxeda.
In a day-long conference here July 22, Intel executives laid out a wide-ranging road map of how they’ll get that done, from new chips that are optimized for particular workloads to innovations like the company’s system-on-a-chip (SoC) methodology to ensuring that Intel has the hardware, software and ecosystem to cover almost every aspect of the data center.
Whatever the system makers and end-users need, Intel will have the products to meet the demand, they said.
“Our goal is that all data center workloads, regardless of what they are, run best on Intel architecture,” said Diane Bryant, senior vice president and general manager of Intel’s Data Center and Connected Systems Group.
Data centers are under pressure to keep up with the growing and changing workloads brought on by the rise of cloud computing, the massive and growing number of devices that are connecting to their networks and the new types of applications they are being asked to run, Bryant and other executives said during the conference. IT infrastructures that are dynamic rather than static, automated rather than manual are necessary, she said.
“We’re going through a fundamental transformation in the way that IT is used,” Bryant said “Today, we look at IT as the service. IT is no longer supporting the business; rather, IT is the business.”
To meet the needs of organizations, Intel is rapidly evolving its products and how the company makes them. Gone are the days when the company would make a general-purpose Xeon chip, and OEMs like Dell and Hewlett-Packard would put them into general-purpose servers to sell to businesses. Now those businesses are demanding products that can help them run data center infrastructures that are automated, flexible, scalable, on-demand and cost-effective.
They need what Intel officials are calling a software-defined infrastructure, where data center resources like compute, storage and networking are pooled, and applications automatically can draw the resources they need to run their workloads, and then return those resources back to the pools for other applications to use.
Intel is moving aggressively to build out its capabilities to help organization meet their needs, in large part with a range of new chips designed to bring higher performance and increasingly lower power consumption. Intel, which is just beginning to see its latest 22-nanometer Haswell architecture find its way into the latest Xeon chips, next year will roll out the 14nm “Broadwell” processors, as well as the first Xeon SoC, which will include such integrated features as I/O, fabric and accelerators for servers, storage capabilities and networking. While offering specifics on power efficiency for Broadwell, Bryant noted that power usage in the last three generations of the lowest-power Xeon E3 has dropped from 20 watts to 13 watts.
At the same time, Intel is expanding its low-power Atom server platform for energy-efficient, dense microservers. The Atom S1200 Centerton chip, released in 2012, will be followed later this year by “Avoton” (for servers) and “Rangeley” (for networking systems)—which will comprise the Atom C2000 lineup—and by the 14nm “Denverton” in 2014. Avoton and Rangeley, which will have up to eight cores, will offer seven times the performance and four times the energy efficiency of Centerton, officials said. The company has been sending out samples to customers for several months, and already has more than 50 systems being designed for them.
Intel Makes Its Case for Data Center Transformation
Intel also will increasingly leverage its massive manufacturing capabilities and its evolving SoC development methodology to create custom chips for individual big customers, Bryant said.
Intel already has made custom chips for such big Web 2.0 companies as eBay and Facebook, said Jason Waxman, vice president and general manager of Intel’s Cloud Platform Group, adding that organizations “are asking us for custom solutions.”
The SoC methodology—which integrates such features as I/O, security and memory onto the silicon—enables Intel to offer products that are more optimized for particular workloads, such as differentiating between systems that run more compute-intensive applications from those that need more networking capabilities. The methodology not only will enable Intel to better develop custom chips, but also to offer a wider range of capabilities in its product portfolios.
In addition, Intel will continue to expand its reach within the data center. The company will grow its portfolios for storage and networks, look to play a significant role in the evolution of such movements as software-defined networking (SDN), and leverage such investments as its acquisitions of Fulcrum Systems and Cray’s fabric technology to expand its networking capabilities.
Intel will bring much of its muscle to its evolving Rack Scale Architecture, with racks that can run server modules powered by Xeon or Atom, and with those modules increasingly sharing resources from power and cooling to I/O, according to Waxman, who showed off prototype “trays” holding three Xeon modules and 30 Atom modules. Eventually, these racks will essentially offer pools of compute, storage and networking resources that can be accessed by applications as needed, he said.
Bryant and Waxman also talked about the need for consistency across an organization’s IT landscape, making the argument that an infrastructure built upon Intel technology—from the chips to the familiar development tools to the huge software ecosystems that are optimized for Intel products—makes more sense than introducing new architectures that require new or recompiled applications.
It’s an argument Intel executives have made when talking about ARM’s planned efforts to make inroads into the data center. ARM officials have countered by noting the increased use of open-source technologies in data centers, and the strong partnerships it has with Linux operating system vendors and other companies in the open-source communities.