Worlds Apart - The nascent threat to the existing data centre market from the Open Compute Project

Last week there were a couple of important conferences cum exhibitions, the sprawling CloudExpo Europe/Data Centre World/Smart IoT/Cloud Security Expo in London’s Docklands, and the Open Compute Project (OCP) Summit in San Francisco. The old world and the new world, literally and figuratively, 5,350 miles apart.

I was at the ExCel in Docklands and was only keeping track on the OCP Summit from afar. But even so, I got a strong sense that what was going on in San Francisco was the shape of things to come. While the Big Data and Smart IoT elements at ExCel had a certain buzz about them, many of the big names in Cloud were not there and for the rest, it felt like people going through the motions.

Now I may be in danger of getting caught up in the enthusiasm of the OCP community. After all, outside the hyperscalers, OCP products have barely registered. IHS Markit research, released at the OCP Summit showed that, leaving out the four OCP end-user Board Members, Facebook, Microsoft, Rackspace and Goldman Sachs, revenue from data centre gear was $1.2 billion in 2017. This is about 1% of the estimated $137 billion data centre hardware market. However, the amount of money, not disclosed, spent by the 4 named Board Members on OCP hardware would increase that percentage significantly.

Moreover, there are strong signs that the telco sector, needing to make large investments in new infrastructure to take advantage of the opportunities offered by 5G, IoT and Edge Computing, are looking to OCP designs to minimise both capital and operating costs. IHS Markit estimates that the “non-board” OCP equipment market will grow at 60% per annum to reach $6 billion by 2021. This could be the start of OCP going mainstream.

Back at ExCel there was little indication of any forward thinking on OCP. There may be people planning for an OCP future at the exhibitors’ HQs, but it certainly wasn’t on show. One server manufacturer positively dismissed them going down an OCP route. I know OCP reduces the ability of vendors to design in their own proprietary widgets and claim levels of performance differentiation, but the attitude smacked of those who, over the years, have stuck to their proprietary guns and subsequently disappeared as the open systems bandwagon rolled over them.

The growth in data and processing requirements that IoT, Big Data and Machine Learning promise, may well disguise the threat that OCP poses to many vendors’ propositions and business models for some years to come. Data Centre building is booming. The need to reduce energy usage is now a critical requirement and data centre equipment manufacturers and associated companies are driving incremental improvements in energy efficiency. But the OCP have design visions and strategies that will seek to remove much of this equipment and change the whole data centre paradigm.

We have seen how Big Software has changed the way we view computer hardware. As an OCP presentation I saw states, “software eats hardware”. Taking this software led approach enables much greater agility and velocity. At Bloor we call that Mutable. I fear that many of the data centre equipment manufacturers and hardware vendors have not yet developed a Mutable Business mindset.