The Open Compute Project (OCP) started an HPC group in August of 2014 to address the needs of the low latency computing enterprise and data center deployments. It is an open group that any individual or company may participate in. While the name of the group is HPC the work done in the group has a vast set of application areas ranging from traditional HPC and supercomputing to public cloud based data center analytics, financial trading and mobile edge computing. The project is focused on three phases designed to create open innovations for the above markets, resulting in specifications and commercially available platforms from the silicon level to the board and sub system to the system level. At a high level some of the objectives of the group include but are not limited by the following:
- Fully open heterogeneous computing, networking and fabric platform
- Optimized for multi-node processor agnostic any to any computing using x86, ARM, PowerPC, FPGA, ASICs, DSP, and GPU silicon on hardware platform
- Enables rapid innovation in low latency high Performance Computing and Big Data analytics through open non-lock-in computing, interconnect, and software stack.
- Energy efficient compute density
- Path to Open Silicon and Open APIs, initially leveraging existing industry standards, later developing its own silicon
- Leverage industry standard interconnects
The project is divided into 3 phases to address gradual roll out/ commercialization of innovation developed in the project over near and medium term time lines ranging from 6 months to 3 years.
During this OCP HPC meet up at CERN on 11 June 2015, many of these topics will be addressed by the presenters and key OCP leaders who will entertain Q&A from European Community collaborators/attendees in attendance.