Five years ago, the Open Compute Project (OCP) was born of a simple idea: Companies could innovate data center technology faster if hardware design were done in the open. Thousands of participants and hundreds of companies have joined in that effort since OCP was founded, and this week they gathered in San Jose to talk open hardware at the Open Compute Project U.S. Summit. This year's conference focused on the evolution of use cases for open hardware, new industries reached through OCP adoption, and challenges to anticipate as we look ahead.

Facebook's Jason Taylor — the current OCP president — opened the conference with a look at how adoption of OCP technologies among highly scaled companies is driving momentum into new organizations and new industries, such as telecom. He also pointed to trends on the horizon, namely 19" data center gear, and opening up the networking stack as the next big step in data center disaggregation. You can watch his keynote here.

Jay Parikh, Facebook's head of engineering and infrastructure, talked about the flexibility that arises as a result of disaggregation — breaking things down into basic building blocks that can then be built up again into different configurations lets companies scale quickly and efficiently. These building blocks enabled Facebook to create new shared experiences like live video and 360 video quickly, and made the infrastructure flexible enough to support advances in artificial intelligence research.

He also emphasized the need to push the boundaries on storage and discussed a new class of technologies using non-volatile memory, including Facebook's collaboration with Intel on improvements in this space. Collaborations are often the driving force behind advancements in the data center, and this year Google announced that it has joined OCP and will work with Facebook on its first contribution, a 48V rack power distribution.

Facebook also announced several new contributions and collaborations at the OCP Summit:

  • Lightning, an NVMe-based storage platform built to improve flash capacity and performance. Facebook is working with Intel to explore this new class of memory technologies.
  • OpenBMC, a low-level board management software that we previously open-sourced to manage our Yosemite server and our Wedge network switch with greater flexibility, will now support storage — including our new NVMe-based storage platform, Lightning.
  • Yosemite, an open server design built for better performance and efficiency by scaling compute power, together with Mono Lake, a server board using a single-socket CPU designed with Intel's 16-core Xeon-D processor.
  • Wedge 100, our 32x100G network switch, and 6-pack, an open modular switch platform.
  • Big Sur, our open AI hardware platform. By leveraging NVIDIA's Tesla Accelerated Computing Platform, we use Big Sur to train machines twice as fast and to explore neural networks twice as large.
  • New optical interconnect standards: Facebook is working with Equinix, Google, Microsoft, and Verizon to help create more alignment around optical interconnect standards using duplex single-mode fiber. Specifically, the focus is on 100G CWDM4 optimized for a low-cost data center environment.
  • For more highlights from the opening keynotes and day 1 of the summit, visit the OCP blog.

    Leave a Reply

    To help personalize content, tailor and measure ads and provide a safer experience, we use cookies. By clicking or navigating the site, you agree to allow our collection of information on and off Facebook through cookies. Learn more, including about available controls: Cookie Policy