Facebook Inc. is stepping up efforts to bring hyperscale efficiency into the enterprise with the release of blueprints for two internally developed systems and the related management software to the Open Compute Project. The launch represents its most significant contribution since starting the initiative in 2011.

The undertaking has since garnered the support of companies in every corner of the industry, from low-cost manufacturers to big names such as Microsoft Corp., which made headlines last year after donating the design for a homegrown server that consumes 40 percent less energy than alternatives. Facebook is now upping the ante with a model of its own that takes efficiently to a whole new level.

Dubbed Yosemite, the design is incorporates up to four small system-on-chip units in a 2U chassis – instead of two beefier processors as conventional machines do – in the form of pluggable modules. Supported cards can be added and replaced as needed regardless of the configuration.

That approach, which is similar to the cartridge system Hewlett-Packard Co. has implemented in its Moonshot line, allows administrators to add processing power on a much smoother curve than one two-processor server at a time. The modules are not only  efficient at a power draw of about 90W but share a single connection as well, which simplifies networking.

The non-standard Open Compute Rack where Yosemite servers are meant to fit in turn connects to the rest of the data center through a homegrown switch that Facebook is also open-sourcing as part of the launch. But it’s not the design itself that is worth paying attention too.

With its single on-board Broadcom Trident II, Wedge doesn’t really stick out from the countless other commodity top-of-rack designs that implement the chip, at least not in the hardware department. The differentiation lies one layer up, in the management plane, where the switch is treated more like a server than a traditional networking device.

The software that Facebook is releasing in conjunction with the design to help manage Wedge clusters implements management functions as district applications. The launch lineup includes a baseboard management controller for aggregating diagnostic data from nodes as well as higher-level administrative an automation services.

That application-based model allows the deployment of new features without adding more complexity to the core platform, which in this case is a local agent installed on each Wedge unit. The separation between the management functionality and the switching core has added advantage of allowing developers to adapt capabilities for different network operating systems.

In particular, Facebook is planning to add support for the Linux-developer management stacks from Cumulus Networks Inc. and Big Networks Inc., integration aimed at saving organizations the hassle of implementing the software completely from scratch and thus lower the adoption barrier. But there is clearly still much more work to be done in order to bring hyperscale computing into the reach of the traditional enterprise.