In the five years since its launch, the Open Compute Project (OCP) has chalked up dozens of innovations and technical specification contributions that have been implemented by hyperscale data center operators. The ambitions have now expanded beyond rack hardware to include switching, storage, silicon photonics, a telemetry framework, an open-source analytics platform and new domain of solutions adapted for telecom operators.
Here are some highlights from this week's OCP Summit in San Jose, California:
- Hundreds of companies and thousands of developers are now participating in Open Compute
- Facebook estimates it has saved several billion dollars thanks to OCP
- The OCP Board of Directors includes Jason Taylor (Facebook), Bill Laing (Microsoft), Don Duet (Goldman Sachs), Mark Roenick (Rackspace), Jason Waxman (Intel), Andy Bechtolsheim, and Frank Frankovsky.
- OCP has big ambitions for the telecom world which needs efficient infrastructure too. OCP has launched a Telecom Infrastructure Project that includes initial participation from AT&T, DT, EE, Equinix, Nokia, SK Telecom and Verizon.
- Facebook is moving quickly to bring 100G technology into the network backbone of its hyperscale data centers
- Intel said its OCP Rack Scale Architecture, with compute shelves, NVMe storage shelves, and network shelves has gained traction from a robust ecosystem of partners.
- OCP rack servers are moving to a standard 19" design. This enables system to pack up to 256GB of DDR4 DIMMs on a single motherboard.
- OCP foresees that specialized workloads, such as security or SDN, will benefit from acceleration boards based on FPGAs.
- Silicon photonics will soon be a requirement in hyperscale data centers, especially when backbones eventually upgrade beyond 100G.
- OCP is developing an open telemetry framework for data centers.
- OCP open telemetry will be matched by an OCP open analytics platform.
- Microsoft highlighted its role in developing an open switch abstraction interface (SAI) to remove complexity in hyperscale data centers
- Microsoft's next OCP contribution is Software for Open Networking in the Cloud (SONiC), which allows sharing of the same software stack across hardware from multiple switch vendors. With its modular architecture and lean stack, SONiC allows data center operators to debug, fix, and test software bugs much faster. It also allows the flexibility to scale down the software and develop features that are required for datacenter and networking needs.
- Artificial Intelligence applications will require highly efficient and massively scalable hardware blocks. Facebook is bringing these AI requirements to OCP. Its Big Sur AI hardware assembly has already deployed thousands of machines in just a few months. Big Sur is Open Rack-compatible and incorporates eight high-performance GPUs of up to 300 watts each, with the flexibility to configure between multiple PCI-e topologies. It uses NVIDIA's Tesla Accelerated Computing Platform.
- Google has just joined OCP. The biggest cloud participants also include Facebook, Microsoft and Rackspace. (still missing Amazon, Alibaba, Apple, LinkedIn, etc).
- Google's first contribution to OCP is its design for a 48v data center rack. Google said its design is 30% more energy efficient, in part by minimizing AC-DC conversions.
- Equinix has announced it is adopting Facebook’s Wedge network switch design and open-source architecture.
- Goldman Sachs announced that more than 80% of servers that the company has acquired since last summer are based on OCP standards.
- OCP is introducing Lightning, which is flexible NVMe JBOF (just a bunch of flash). It is designed to provide a PCIe gen 3 connection from end to end (CPU to SSD). It leverages the existing Open Vault (Knox) SAS JBOD infrastructure to provide a faster time to market, maintain a common look and feel, enable modularity in PCIe switch solutions, and enable flexibility in SSD form factors.
http://www.opencompute.org/
