ARM Server Chips Get Xen Virtualization Support
Proponents of low-power ARM-based servers got a boost this week with the latest release of the open-source Xen virtualization hypervisor, which offers preview support for the ARM architecture.
The Linux Foundation, which in April took over the lead on Xen development from Citrix Systems, on July 9 released Xen 4.3, which included several key enhancements and capabilities. Among those was what the foundation called "experimental support for ARM virtualization."
This support not only includes the current 32-bit variant, but also the upcoming 64-bit capabilities that will arrive with the ARMv8 architecture. George Dunlap, a software engineer with Citrix, said in a post on the Xenproject.org blog that the 32-bit port of Xen boots on the ARMv7 architecture, which includes the Cortex-A15 platform. On this architecture, Xen can create other virtual machines and "supports all the basic virtual machine lifecycle operations."
While hardware for 64-bit ARM systems-on-a-chip (SoCs) is not yet available, engineers have Xen "running well in 64-bit mode on ARMv8 Real-time System Models by ARM," Dunlap wrote.
ARM's low-power SoCs dominate in the booming mobile device space, where the bulk of tablets and smartphones run on ARM-designed chips made by the likes of Samsung, Qualcomm and Texas Instruments.
However, executives with ARM and its partners, such as Calxeda, Marvell Technology and Applied Micro, see an opportunity to move into the data center with the growing demand for dense, low-power microservers from Web 2.0 companies and cloud providers, such as Google, Facebook and Amazon. Such companies are looking for systems that can run efficiently in their massive data centers and can process huge numbers of small workloads.
The move into servers will open up another area of competition with Intel, which dominates the global server chip market and is addressing the microserver space with its low-power x86-based Atom platform. The company last year launched its Centerton chip, and later this year will roll out the next-generation Atom chip aimed at the space, dubbed "Avoton."
ARM's current 32-bit architecture lacks some key capabilities needed for server chips, including 64-bit computing and strong virtualization. However, such capabilities will come with ARMv8, which ARM officials said will be appearing in systems next year.
ARM and its partners believe there is room in the server space for the architecture. Advanced Micro Devices, which for decades has been competing with the larger Intel in the x86 server chip space, has released an x86-based Opteron server chip—dubbed Kytoto—for the microserver space. At the same time, the vendor also will develop its own ARM-based server chips starting next year.
Andrew Feldman, corporate vice president and general manager of AMD's Server Business unit, told eWEEK in May that he believes ARM could claim as much as 20 percent of the server chip market by 2016. The changing data center environment—from the rise of cloud computing to the rapid growth of mobile devices to demands of Web 2.0 companies—is fueling the demand for an architecture like ARM's, Feldman said.
"This new environment is going to have new needs, and the same-old, same-old will not work anymore," he said.
Lakshmi Mandyam, director of ARM's Server and Ecosystems unit, said in April that the partner ecosystem around the company—from chip makers to software developers—will be key in driving the company's efforts in the server space. Mandyam also said the growing use of open-source technology in data centers also will be an advantage for ARM.
"Open source is the great equalizer," she told eWEEK. "I don't think the gap [between ARM and Intel in server processor technology] is as much as you might think."
Some top-tier server makers, including Hewlett-Packard and Dell, also are looking to ARM and its partners for their microserver efforts. HP in April announced the first of its low-power Project Moonshot systems, though it is being powered by Intel Atom SoCs. However, ARM-based chips will be used in models coming out later in the year.