Open Compute Project – Rising Relevance And More Stakeholders

Background

Today’s announcements at the Open Compute Project (OCP) 2013 Summit could be considered as tangible markers for the OCP crossing the line into real relevance as an important influence on emerging hyper-scale and cloud computing as well as having a potential bleed-through into the world of enterprise data centers and computing. This is obviously a subjective viewpoint – there is no objective standard for relevance, only post-facto recognition that something was important or not. But in this case I’m going to stick my neck out and predict that OCP will have some influence and will be a sticky presence in the industry for many years.

Even if their specs (which look generally quite good) do not get picked up verbatim, they will act as an influence on major vendors who will, much like the auto industry in the 1970s, get the message that there is a market for economical “low-frills” alternatives.

Major OCP Initiatives

To date, OCP has announced a number of useful hardware specifications, including:

• Common motherboard specs for both Intel and AMD compute nodes – At the summit, AMD announced the AMD Open 3.0 platform (formerly code named “Roadrunner,”), a dual socket server suitable for a wide range of workloads, including enterprise workloads as well as web server, Java, and other web workloads.[i] The latest spin on a common motherboard, in addition to the base Intel and AMD reference architectures is the “Group Hug” design, which specifies a common slot design for plug-in CPU cards. Currently the Group Hug specification supports AMD, Intel, and ARM CPUs in its common slot. The impact of this board is unclear – since it is a PCIe connection, the common boards will have to contain all their own memory and will have a long latency (relative to memory operations) round trip from the card to any other system resources. This spec is most likely to be a place keeper for emerging SOC technology from ARM vendors and AMD and Intel, as they package complete multi-node SOC implementations into form factors that will fit the common slot specification.
• Common rack specifications for both a customized (think Facebook) form factor and a more standard rack specification for 19" form factor devices – In addition to the existing OCP rack specifications, Intel and Facebook on Wednesday announced a new rack architecture designed to allow disaggregation and scaling of storage and servers with 100 Gb/sec low-cost silicon photonics components that have been under development by Intel for many years. Think multi-year adoption curve but very forward-looking architecture for large data centers. It is almost certain that this announcement is a prelude to Intel commercializing this technology through other OEMs and partners.
• Open Vault – A specification for a high-density SAS storage module, capable of packing 30 drives into a 2U form factor, a very useful building block for many applications that need lots of low-cost storage and can live without the last quanta of performance that FC SCSI delivers.

In addition, OCP has announced a more ambitious strategy of “grid to gates,” echoing several vendor programs to look at the entire compute stack from low-end silicon to entire data center scale. How successful OCP’s higher and lower layers of their strategy will be will depend on how much traction they can get with their core server, rack and storage module specifications.

My initial impression is that the server and storage modules (especially if they add a SAN-connected variant) along with their rack and power architectures have “legs” and will get adopted by multiple second-tier vendors and will also strongly influence the efforts of the major systems vendors. Currently, the list of partners and, more importantly, active participants, is approaching critical mass, particularly since they have the backing of AMD and Intel, have at least one Mega-Customer (Facebook), and are actively embracing emerging technologies such as ARM. RackSpace has already stated that they may make OCP-compatible servers, and I would not be surprised to see others following suit.[ii]

Their management initiatives and macro-scale data center efforts are more debatable, since the ecosystems and channels for these are more complex and involve more legacy investment on the enterprise side, but appear essentially sound. Even if not adopted verbatim they may be a powerful influence on the industry.

Who Benefits And Who May Suffer?

At a high-level, the obvious beneficiaries are the extended server purchasing community. Everything about the OCP is designed to drive costs down, and while all users will not elect to follow this minimalist compute utility curve, it clearly exposes the extra costs being added by mainstream server vendors, which will have two major effects. First, it will force these vendors to selectively adopt some of the OCP design practices into their products (some have already started producing minimalist servers; others will follow). It will also force the mainstream vendors to continually innovate and add opex-reducing value to their servers in order to justify the increased capex. We have already seen this trend with the latest set of releases from Dell, HP, and IBM, all of whom have added significant software value to their servers to simplify both initial installation and ongoing maintenance. Vendors who do not clearly differentiate on cost or value-added services will be in trouble.

This post is essentially an early stake in the ground; several of the new initiatives and technologies announced at this summit are in themselves blog-worthy, and I expect to see considerable activity from established and new partners going forward.