AMS-IX has expanded its EasyAccess footprint to the maincubes datacenter in…
So what is Open Compute anyway? (And why might OCP be relevant to you?)
In a previous blog about Open Compute Project (OCP) I indicated that maincubes, together with our partners Rittal and Circle B, has opened a so-called OCP Experience Center in our data center in Amsterdam. But what exactly is OCP?
Open source hardware
Open Compute Project is probably the best known community when it comes to open source hardware. For the start of the project we have to go back to 2009. That year Facebook grew extremely fast. Millions of people were already using the social media platform and that number grew exponentially. Only an excellent data center infrastructure is able to support such growth. Within Facebook people were becoming aware that the technology and the systems supplied by traditional IT suppliers were simply unable to support this growth. The existing solutions could not keep up with the scale of the IT infrastructure Facebook needed. They were not robust enough, consumed too much energy, offered features Facebook did not need nor wanted. So what to do?
Develop your own data center
The social media giant decided to free up a small team of engineers. Their assignment was to develop the world’s most energy-efficient data center, which was also technically capable of facilitating the very substantial growth in the use of the platform. And all this at the lowest possible cost. The team engineered from scratch everything a data center needs: servers, racks, power supplies, switching, cooling systems, management software and much more. The first data center based on these new designs is Facebook’s facility in Prineville, Oregon.
This data center uses 38% less energy than a state-of-the-art traditional data center at the time and was 24% cheaper to run. The main reason for these results is innovation, Facebook realized. But how do you keep this innovation going?
Open source designs
Facebook already had a lot of experience with open source software. Would the same community-based form of collaboration also work when it comes to data center hardware? More and more Facebook became convinced that – yes open source hardware might be as successful as open source software.
That’s why in 2011 the company decided to publish its internally developed hardware designs. At the same time the company launched the Open Compute Project – together with Rackspace, Goldman Sachs and Andy Bechtolsheim. The latter is known as one of the founders of companies like Sun Microsystems and Arista Networks.
Open source hardware projects work in a similar way to open source software projects. The community manages an ever-increasing number of projects that develop designs for specific use cases. Think of fairly broad themes such as racks or servers. But there are also projects that look at much more specific topics, such as firmware. Companies are allowed to become members. The same applies to a certain extent to individuals. Through this membership they have access to all designs. These designs can be downloaded and used to manufacture products based on those designs. Or you can develop an extension or an add-on to a design. You can also participate in various projects to influence designs. And for those who have a completely new idea for a project there are also plenty of possibilities. A fairly recent example of such a new project is liquid cooling. Members are required to publish any add-ons or modifications to designs back to the community again.
This is very relevant
Why is this relevant to you? The key word here is ‘innovation’. Innovative solutions for data centers and IT infrastructures offer major advantages – for example when it comes to performance. Or because the costs of running certain IT workloads can be drastically reduced. Or because you can handle a lot more workloads with far fewer operations staff. Or you can finally use the blazing fast hardware you need for your machine learning applications.
Adoption grows fast
Open Compute hardware is no longer just for Facebook. Hyperscalers like Google and Microsoft are also making massive use of OCP hardware. The same goes for large banks, telecom operators and other enterprise organizations. The combination of lightning-fast performance, superior availability, significantly lower energy consumption and much lower management costs is proving to be irresistible to many organizations, as research by IHS Markit shows. Their analysts have calculated that the market for OCP solutions is growing rapidly. While $1.16 billion worth of OCP hardware was purchased in 2017 by parties that do not belong to the Open Compute Project, this amount is expected to reach $10.7 billion by 2022. That is an average growth rate of over 56% per year.
How to train your staff?
This is all good and well – you may say – but my IT people have no experience whatsoever with anything other than traditional hardware and IT solutions. We understand that. And that’s exactly why maincubes now has an OCP Experience Center available in Amsterdam. Here your IT and datacenter staff can see (and touch!) OCP hardware for the first time. They can use the center to experiment with the many solutions OCP has to offer. And once they have engineered their own OCP based hardware the OCP Experience Center is also available to take these solutions for a test drive. And once you are ready to deploy your first OCP based solution maincubes is able to host these for you.
Click here to find out more about the OCP Experience Center. Or watch this short video.
Stephan Harren is Director of Operations at maincubes