The Open Compute Summit is underway in San Jose, California, and today, Facebook announced new contributions to OCP that create more efficient and sustainable infrastructure worldwide. Over the past four years, Facebook has been working to change data center hardware. The company started with new server designs, power handling, and cooling; then focused on storage and rack; and over the past year, it has completely opened the data center network. The result is that today the social network has open-sourced every major physical component of its data center stack - a stack that is powerful enough to connect 1.39 billion people around the world, and is efficient enough to have saved us $2 billion in infrastructure costs over the last three years.

Facebook today unveiled the next steps related to the Open Compute Projec.

The company introduced "Yosemite", a new SoC compute server that increases speed and more efficiently serves Facebook traffic. Developed by Intel, Yosemite is Facebook's first system-on-a-chip compute server that supports four independent servers at a performance-per-watt superior to traditional data center servers for heavily parallelizable workloads.

For Yosemite, Facebook defined each server node as a pluggable module. Each module holds one SoC targeting up to 65W TDP, multiple memory channels with standard DDR DIMM slots, at least one local SSD interface, and a local management controller. Facebook also standardized the module interface such that compliant cards and systems can interoperate. The Yosemite system holds four SoC cards consuming up to 400W total power, which provides about 90W for each SoC card.

Diving into the Yosemite design, the following design elements are important to the system:

• A server-class SoC with multiple memory channels, which provides high-performance computing in 65W TDP for SoC and 90W for the whole server card.
• A standard SoC card interface to provide a CPU-agnostic system interface.
• A platform-agnostic system management solution to manage the system and these 4 SoC server cards, regardless of vendor.
• A multi-host network interconnect card following OCP Mezzanine Card 2.0 specification, which connects up to 4 SoC server cards through single Ethernet port.
• A cost-effective, flexible, and easy-to-service system structure.

This system will be fully compatible with Open Rack, which can accommodate up to 192 SoC server cards in a single rack. Mellanox has already enabled the multi-host support in its next generation ConnectX-4 OCP Mezzanine Card.

Facebook has also proposed the contribution for the Wedge spec, its top-of-rack network switch to the OCP Foundation. And to help make it easier for people to buy and start deploying this open top-of-rack switch, Facebook is working with Accton, Broadcom, Cumulus, and Big Switch to create a Wedge product package for the OCP community. Accton will begin shipping Wedge in the first half of 2015.

The company also released OpenBMC — open, low-level board management software that enables flexibility and speed in feature development for BMC chips. OpenBMC provides an open software framework for next-generation system management. Wedge will be the first hardware powered by OpenBMC, and 6-pack is next.

Facebook also opened the central library of FBOSS – the software created for its top-of-rack network switch, Wedge. The FBOSS Agent is built on Broadcom’s OpenNSL library to program the Broadcom ASIC inside Wedge.

Thanks to OCP and related efficiency work, Facebook said it has saved $2 billion in infrastructure costs over the course of the last three years. And in the last year alone, the company has saved enough energy to power nearly 80,000 homes. The carbon savings associated with that energy efficiency are equivalent to taking 95,000 cars off the road.