AMD Upgrades Open Compute Project Servers

AMD officials are saying servers based on the company’s Open 3.0 specification offer the same performance as proprietary systems from competitors but at less than half the cost.

AMD announced on 13 May that systems based on the Open 3.0 server motherboard – once code-named “Roadrunner” – are now available from such partners as Penguin Computing, ZT Systems and Avnet.

Open Compute Project

The spec is part of the larger Open Compute Project (OCP) started by Facebook in 2011 to encourage vendors to create open-source standards for highly energy-efficient data centres and IT hardware. Through the new platform, AMD officials are looking to make it easier for IT professionals to customise their systems based on workload needs.

The new systems come a year after AMD unveiled its Open specification. Five months ago, AMD officials introduced the Open 3.0 server motherboard, which complies with requirements laid out by the OCP and gives organisations greater flexibility in their server infrastructures and data centre operations.

Now the three partners named above, as well as Hyve, will offer servers based on the Open 3.0 motherboards. The result will be servers with high performance, good energy efficiency and lower cost of ownership, according to AMD officials.

Pointing to benchmark comparisons, AMD officials said that in a virtual desktop infrastructure (VDI), Open 3.0-based servers can run as many virtual desktops as proprietary OEM systems while reducing costs by 57 percent – $4,589 (£3,000) for a single Open 3.0 server compared with $10,669 for a proprietary system.That essentially reduces the cost per virtual desktop from $91.19 to $38.24.

“Global IT organisations have the difficult task of choosing between price and performance when investing in servers,” Bob Ogrey, cloud evangelist and fellow at AMD, said in a statement. “We don’t believe organisations have to compromise one for the other.”

Opteron 6300 chips

The Open 2.0 systems are powered by AMD’s Opteron 6300 chips and can be installed in standard 19-inch racks without modification or in Open Rack environments, according to the company. They’re managed using such open standards as Systems Management Architecture for Server Hardware (SMASH), which comes from the Distributed Management Task Force.

The system motherboards, which are 16 inches by 16.7 inches, can fit into 1U (1.75-inch), 1.5U (2.6-inch), 2U (3.5-inch) and 3U (5.25-inch) rack servers, and hold two Opteron processors, each of which features 12 memory sockets that allow up to 385GB of DDR3 memory.

They also feature six Serial ATA connections per board, one dual-channel Gigabit Ethernet network interface card (NIC) with integrated management, up to four PIC Express slots, one serial port and two USB ports. There also is a mezzanine connector for compatible cards from such vendors as Mellanox (for I/O) and Broadcom (for management).

Web 2.0 companies such as Facebook, Google and Amazon run huge, highly dense data centres full of low-power servers that can run massive numbers of smaller workloads.

Given their data centre environments, such companies are always looking for high-performing, highly energy-efficient smaller servers, and in many instances have resorted to creating their own systems using off-the-shelf technologies. Facebook created the Open Compute Project after open-sourcing its own system designs, and over the past two years it has grown to more than 50 official members.

Facebook specifications

The OCP was developed to encourage technology vendors to produce new data centre products based on Facebook’s specifications, which the company and other organisations could use in their data centres.

The OCP has branched out from focusing on servers to include other data resources such as storage appliances and power supplies.

OCP officials said earlier this month that the consortium is now turning its attention to data centre networks, focusing on such trends as software-defined networking (SDN) and network-function virtualisation (NFV).

Penguin Computing officials announced on 8 May that OCP officials had named the company an OCP solutions provider. Penguin now can offer highly integrated, scalable, custom-built OCP solutions aimed at medium-sized and large data centres, company officials said.

In a post on the company blog announcing the designation, Arend Dittmer, director of product management at Penguin, wrote that while the introduction of open standards could lead to more commoditisation of data centre hardware, there will continue to be demand for vendors with the skills to integrate the open hardware and software, particularly in such complex areas as high-performance computing (HPC).

“So on the one hand further ‘commoditisation’ will add pressure on hardware prices … no doubt about that,” Dittmer wrote. “On the other hand, lower prices drive more and larger deployments with inherently increasing complexity. While there are numerous providers that know how to build servers, the air gets thinner when it comes to solution providers that really know how to make things work.”

HPC, VDI, cloud computing

AMD officials said Open 3.0-based systems are aimed at such environments as HPC, VDI deployments and cloud computing.

“This type of model … is what a lot of large customers and data centres are looking at,” Ogrey told eWEEK in January.

AMD, Broadcom, Quanta and other partners are scheduled to discuss the Open 3.0 platform during a roundtable discussion on 14 May in New York City.

Do you know all about UK tech leader ARM Holdings? Take our quiz!

Originally published on eWeek.