Facebook’s FB artificial intelligence, or AI, research unit has already open-sourced most of its code and research. Now for the first time, Facebook is sharing its own AI hardware design. The Menlo Park, Calif.-based company is open-sourcing the design of its newest server for large-scale artificial intelligence (AI) computing, which it plans to submit to the Open Compute Project, which shares designs of data center products among a number of large companies.
There is a misconception out there that the hyperscalers of the world are so rich that they always have the shiniest new toys in their datacenters. This is not correct. Hyperscalers are restrained by the same laws of physics and the same rules of accounting and budgeting as the rest of the IT sector.
As the theme goes this year, what’s old is new again. And not just in terms of technologies like deep learning or FPGAs, which are suddenly returning to life with new vigor. The trend appears to suit select companies as well. Consider, for instance, Penguin Computing, which opened its doors in 1998 to kickstart the burgeoning Linux server space with the hopes of tracking down customers in the scale-out enterprise world as well as the emerging dot com companies.
Nov. 10 — Penguin Computing, a provider of high-performance computing, enterprise data center and cloud solutions, today announced the deployment of YJ America’s OpenStack infrastructure on Penguin Computing’s Relion OCP servers for customer Yahoo! JAPAN. The adoption by Yahoo! JAPAN with Penguin Computing’s partner ITOCHU Techno-Solutions America, Inc. validates Penguin Computing’s global reach, partnership focus and ability to address the largest and most complex workloads in computing.
Mellanox Technologies, Ltd. (NASDAQ: MLNX) says the Open Compute Project (OCP) has accepted a pair of its 10 Gigabit Ethernet/40 Gigabit Ethernet switch systems. As its name implies, the Facebook-inspired OCP seeks to promote the collaborative development of open source software and hardware for data center, computing, and related applications. Vendors can submit their products for acceptance into the program.
In the first publicly disclosed deployment by a government agency of computing hardware based on specs that came out of the Open Compute Project, the Facebook-led open source hardware and data center design initiative, the US Department of Energy has contracted Penguin Computing to install a supercomputing systems at three national labs.
At the OCP Engineering Workshop, Accton Technology open sourced the design of the AS7500-32X 100 Gigabit Ethernet open network switch, making it an early entry of open network switches based on the Cavium XPliant Switch ASIC. As far as the packaging and x86-based CPU are concerned, the AS7500-32X is the same as the AS7700-32X 100 GbE switch that Accton contributed to the Open Compute Project in March. The AS7500 also uses the same designs for power supplies, fans and enclosures – all in line with Accton’s strategy to develop a consistent open networking switch line.
Open Network Linux (ONL) Unifies Open Compute Project (OCP) Network Operating System Stack, Accelerates Adoption By Both Users And Developers
SANTA CLARA, CA–(Marketwired – Oct 8, 2015) – The Open Compute Project has helped Facebook to save more than $2 Billion, and has become a focal point of open initiatives for data center infrastructure. While numerous open source networking components have been contributed to the Open Compute Project’s networking track, on October 9th a joint demo from Big Switch Networks, Facebook and NTT will mark the first integration of these disparate components into a unified, open source switch operating system: Open Network Linux http://opennetlinux.org.
Today, HP officially launches the open source Open Switch network operating system. Intended to be the base of an open community, the new effort is supported by VMware, Intel, Broadcom, Broadcom and Arista. Mark Carroll, Vice-President and CTO at HP, emphasized that the OpenSwitch operating system is really a community-driven development base. Though HP is announcing the OpenSwitch OS at the Linuxcon EU conference, the effort is not formally backed by the Linux Foundation.
Microsoft has built a Linux-based cross-platform operating system for running network devices like datacenter switches. ACS is built around the Open Compute Project’s Switch Abstraction Interface (SAI), which is the standards C application programming interface for programming ASICs. It’s software for running network devices like switches. The Open Compute Project accepted the SAI specification — contributed by Microsoft and other contributors including Mellanox, Dell, Broadcom, Cavium, Barefoot, and Metaswitch — in July 2015.
Jason Taylor, vice president of infrastructure at Facebook, was elected as the new chairman of the Open Compute Project (OCP) at the organization’s meeting in New York last week. He replaced Frank Frankovsky, one of the original architects of Facebook’s Open Compute. The appointment was announced in blog post earlier today. The OCP Foundation board of directors meets every quarter. Once every year these meetings include the nomination and election of board leadership for the upcoming year.
AMAX Helps Financial Services Market Converting Data Centers to OCP Wall Street Rethinking Data Center Hardware
AMAX Centurion is based on the Decathlete spec, but with the option of being configured with NVMe flash storage, high speed networking and high performance CPUs for a risk analysis, financial forecasting and modeling platform, as well as serving as an OpenStack and Hadoop platform. At 2015 OCP Summit, Bank of America along with Capital One and JPMorgan Chase announced their intentions of adopting OCP into their own data centers.
Intel RSA is the chip vendor’s own modular architecture design that disaggregates compute, storage and networking and weaves them together in a fabric it claims makes resources easier to pool and pod. Now Quanta has developed a proof of concept for a server that blends Intel’s RSA specs and Open Compute designs. The hardware vendor, which already offers hardware based on Open Compute designs, claims will significantly reduce datacentre energy consumption and costs, reduce vendor lock-in and ease management and maintenance.
Facebook today is announcing that it has expanded the capabilities of OpenBMC, the open-source software that the social networking company developed to help it control certain aspects of its data center hardware. When Facebook introduced OpenBMC — which is essentially a full-on Linux distribution to deal with hardware components such as sensors — in March, the company said the tool could controlFacebook’s Wedge custom networking switches. But Facebook has since gone further and added support for Facebook’s efficient Yosemite servers.
Software developers obviously love open source. They get to collaborate, build on top of work already done by others instead of constantly building from scratch, and add features they need to existing solutions. Innovation often happens faster in open source communities than it does behind closed doors of corporate development departments. While software runs in data centers, data center managers’ job doesn’t usually extend much beyond making sure there is enough IT, power, and cooling capacity to support the application workloads and making sure the systems are configured and secured properly. But it isn’t going to stay this way forever.