Strategic partnership around Open Compute technology

We work mainly with hyper scalers. IT infrastructure companies (servers, storage, network...) are both a growth lever and a cost where the carbon footprint is too important.

3-4 years ago, these players asked us to find innovative solutions in order to gain scalability, to consume less, to slip from computer manufacturer control. We then became interested in the Open Compute Project, which is the computer hardware designed by the Gafams more than ten years ago, with a twofold objective:

• reduce investments
• reduce the environmental footprint by consuming less and creating machines that have greater durability than what was available on the market.

We positioned ourselves on this market. It is now part of our activity and allows us, precisely because the machines designed by Gafam are more resistant, to have a slightly different supply channel and to be able to offer recertified machines.
When you buy a server, 70% of its carbon footprint is generated during its design. The best energy saving is the one we don't consume. Open Computing perfectly meets the constraints.

In this mindset, we discuss with Qarnot since they use these machines in their solutions.

At Celeris, our customers create meaning thanks to the infrastructure, and this infrastructure is also a cost. They have significant server farms and their objective is to reduce the initial investment and the OPEX: power consumption, maintainability, being more independent, and with more open hardware.

3-4 years ago, these same customers asked us: \"how can we reduce consumption, not necessarily change machines every 3, 4 or 5 years and invest in this path?\"

It is a real change in the last two years. The meetings we attend, we see a lot of CSR departments appearing with quite drastic carbon footprint reduction objectives over the next 10 or 15 years.

There is not much of a quick fix. The GAFAMs asked themselves the same questions long before us, about 10 years ago, perhaps driven more by the potential financial savings to be made than by the carbon footprint. But it turns out that today, we are bringing together the best of all worlds with open computing, namely materials that have been designed to last easily for 10 years and that can be repaired very easily.

We have a test lab with the materials proposed by the standard players in the market but also open compute racks and we benchmark all of this. We try to see what holds up best, what consumes the least, etc. and it's true that with an open, compute server, in 1 minute 30, you can change a motherboard, there is no tool, it's not proprietary, it's open source. Everybody can build these machines, the plans are available, it's important for the customers.

So today we are recertifying machines that are 3-4 years old and have nothing to be ashamed of compared to other manufacturers' machines.

Open source was made for security so that the community could improve things and have visibility, and above all the Gafams had servers with components that were useless to them. I wouldn't say that Open Compute is custom-made, but they have rethought computing. They wanted to start from scratch. So they transformed the racks and based the power supply in the rack rather than in the machines because it's better to have one big power supply than lots of small ones,

They made the machines more airier with fewer components inside. All the cabling is done from the front so the air circulation is from the back with better heat restitution.

In summary I would say that they designed the machines that suited them, that is to say easy to get, we know what's inside, which consume less and last longer.

It's the design of the whole solution, especially the compute in Qarnot, which means that we will consume much less (about 30% less on average). We even have security customers who have done benchmarks, it's obvious.

In France, we are a bit late. In the Netherlands, they have already adopted OCP to heat greenhouses, etc. When I discovered Qarnot two years ago, I got in touch with Paul who explained the concept of transforming servers to recover heat, running pipes through the servers, etc. And when you go to HP, Dell or Lenovo, it's a bit complicated to open up and change everything in terms of the warranty. It's just not meant to be.

So the OCP that is open and not locked to a manufacturer is what Qarnot needed.

We did some hardware tests, Qarnot's engineers worked on it and we have to admit that it's a winning bet.

On the environmental side, we have to move. We think that by digitizing and sending to the cloud, we consume less, but it's getting worse and worse. Data centers are an enormous source of pollution. They sometimes consume as much electricity as a small town. So we don't have a choice, we have to find solutions that allow us to pollute the planet less.