Critical Thinking is a weekly column on innovation in data center infrastructure design and management. More about the column and the author here.
Since the Open Compute Project was founded in 2011 the balance of power in the data center industry has continued to shift.
True, most of the world’s data centers – about four million of them, according to some estimates – are still owned and operated by enterprises. But the bulk of those are small IT rooms or closets. Most large cloud service providers still use servers or other gear from traditional technology suppliers but have also seized the initiative by sourcing custom hardware from ODMs as well as developing and deploying OCP equipment.
There are now more than 390 hyper-scale data centers in the world, and that figure is set to rise to more than 500 by 2020, according to the latest estimates by Synergy Research Group.
The hyper-scalers’ purchasing power and ability to innovate were writ large with Hewlett Packard Enterprise’s decision last year to back out of selling its commodity Cloudline hardware. The supplier deemed the business no longer financially viable.
But the shifting of the center of gravity toward hyper-scalers using open OCP architectures – AWS being the exception – raises questions about OCPs long-term relevance. In its early days, its then chairman and president Frank Frankovsky talked about disrupting the traditional supplier-led data center hardware industry, which was in some cases stifling innovation.
"We're building something cool that's unique to the industry," he said back in 2013. "This is a consumer-led organization. I think suppliers are starting to listen more. We're designing and delivering in the open. By doing this in the open, we're creating a virtuous cycle."
While OCP gear is still not widely adopted outside hyper-scale data centers – 451 Research estimates that open source infrastructure currently represents less than 5 percent of the total data center spending globally – it is no longer the upstart it was once positioned as (despite the presence of Intel, Goldman Sachs, and Facebook amongst its founders). And if OCP eventually has less to push back against, then its army of volunteers may be less motivated to donate their time gratis.
However, it’s probably fair to say that these existential questions won’t become practical problems for the organization in the near term. For now, OCP is maintaining its momentum by continuing to push at its boundaries.
Speaking to Data Center Knowledge, Steve Helvie, VP of channel at the non-profit organization, said OCP is targeting several key markets in 2018. These include telcos, service providers (from SaaS to colocation), financial services (including blockchain), high-performance computing, healthcare, and government.
Part of Helvie’s remit is trying to convince more colocation providers to design facilities and provide services that make it easier for enterprise customers to deploy OCP equipment. Large operators such as Equinix and smaller players such as UK-based colo Aegis are already onboard. “I do see a massive momentum in Europe for the colo space,” he said.
To that end, the organization released a set of guidelines (via the OCP Wiki) for colocation providers at the end of the year. It has also developed a checklist to help verify that colo sites are appropriate for OCP deployments.
A report by 451 released last year revealed that 51 percent of customers believed that the availability of OCP-ready white space would make them more willing to use a colocation provider.
The end game, which has been discussed for a couple of years now, is some form of OCP certification for colocation providers. “We are currently evaluating how best to recognize and brand colos that are OCP-ready,” said Helvie.
Initial adoption of any OCP colocation cert is expected to happen in Europe, as this where most of the effort has emanated from. “All of the guys in Europe were extremely active in putting together those guidelines, which is why I am more excited about the European market than I am about the others right now,” said Helvie.
However, the exact form the OCP-ready stamp will take is still being developed. “We are likely not going to have another brand, but it will be a level of formal recognition. I want enterprises to be able to go into our marketplace and say, ‘Where can I find someone who is ready to host Open Compute?’”
As well as providing OCP-ready space – perhaps without centralized UPSs and lighter redundancy overall than traditional colo space – colo providers may also support customers with sourcing and procuring OCP equipment – something that has been a challenge for smaller organizations.
Specialist system integrators, or SIs, will probably play a key role in that process alongside the colocation providers, according to Helvie. “I see a huge need in Europe and Asia for local SIs. Currently, many businesses are still evaluating how best to approach building a business around OCP.”
Another important shift for OCP in the coming year could be greater involvement in software and applications – something it has shied away from in the past. “This year we are going to be announcing ways in which companies can contribute more than just a hardware design. That could be everything from reference architecture to a software validation package. We are broadening our ability to welcome new members.”
But OCP doesn’t have everything its own way. As well as continued competition from OEMs and the absence of AWS, another open source data center hardware initiative – Open19 – emerged in 2017. Led by LinkedIn, Open19 is targeting smaller companies and edge deployments. It also has looser licensing terms, and it’s still not entirely clear how to the open hardware groups will coexist.
Another challenge for OCP is to ensure that it retains its original open source ethos and doesn’t just become a marketing rubber stamp for the large suppliers and service providers that have joined its ranks over the years. Helvie believes the organization’s structure will protect against this. “If you are a tiered member, one of the top organizations, you have got to contribute, and those contributions will ensure that it remains more than just a rubber stamp.”
Ultimately, while OCP will continue to push into new verticals and layers of the technology stack, it is probably an organization with finite raison d'être. The rate at which it can convince smaller operators to adopt OCP equipment is unlikely to outpace the speed at which enterprises are shifting workloads to public cloud providers, some of whom are already OCP converts. (Microsoft has standardized on OCP across its entire global cloud infrastructure.)
And while AWS may not be a member, it has also very obviously seized some of the innovation initiative from traditional OEMs in line with OCP’s original mandate.
When (not if) public clouds running open or proprietary custom hardware become the default option for most enterprise workloads, OCP will have achieved much of its original mandate. But it may also find its continued relevancy in question, as will a large number of other organizations.