For Day Two of the fifth edition of Open Compute Summit in San Jose, California, Cole Crawford (Executive Director with OCP Foundation) was joined on stage by Justin Erenkrantz (Head of Cloud Architecture, Bloomberg), Jeremy Huylebroeck (Software and IT Architect, Orange) and Ron Williams (VP of Operations, Riot Games) for a Keynote panel exploring the adoption of OCP technologies.

As they were all representing super early adopters of Open Compute, Crawford wanted to know exactly what led them to embrace this technology.

 .

• Driving down costs + improving energy consumption

First one to answer, the VP of Operations with Riot Games, explained:

“As we continued to grow and pick up more and more players and engaging with products, that was driving the infrastructure scale and we started looking at some of the really large web-scale guys. It was perfect timing: Facebook started rolling out open compute offering a glimpse into how they were doing huge web-scale implementation. We adopted it very fast because we’re always looking for more ways to return more value in what we’re doing to the players – driving down the costs and improving the energy consumption,” said Ron Williams.

 .

• Energy efficiency + commodity hardware

“Being part of the Orange Corporate office in San Francisco, we always look at the trends that could bring value to Orange as a global company. In the past five years, Orange has been working a lot on energy-efficiency for the data centers and also commodity hardware. What Facebook did resonated very well with us, and we wanted to adopt this technology very early, even if we knew we would not go to production right away,” explained Jeremy Huylebroeck.

 .

• The time was right

“We had someone from our data center team here at last year’s summit and he came back talking about OCP. We heard about it, but we felt now was really the time where we could have all the supply chain and the designs to take that jump,” added Justin Erenkrantz, Bloomberg’s Head of Cloud Architecture.

“Mark Zuckerberg has recently talked about Facebook in the ’90s and what Capex investment they would have had to make if they started the company 20 years ago,” noted Cole Crawford. “Riot Games has deployed a very large game, responsible to up to three percent of all daily internet traffic. How do you weigh the Capex investment versus the ability to work backwards?” he asked.

 .

• Open Compute lowering the power usage, Capex volume

“Microsoft had a pretty cool slide yesterday, talking about how you move from 100 to 1,000 to maybe 10,000+ servers. The considerations you have to worry about are maintainability, the return of service, the workload in the data center. Open Compute has helped out with some of those. Once we simplified what we were doing, it was pretty simple math: 30-40 percent less power usage, the Capex volume was significantly cheaper as we went buying with discounts from our suppliers,” answered Williams.

“What does Orange think of the way you can adopt OCP?” inquired Crawford.

 .

• OCP is the enabler in facing the fierce competition

“Our main goal is optimizing some of our cloud services today; it could have a bigger impact at Orange in the enterprise marlet, but it’s too early for that. The solution we have in France is based on OpenStack and we see Open Compute as an extension of that philosophy where the OpenStack and the open source layers allow us to be very agile.

“Hardware has to play a role in that, first by reducing the costs;the open source hardware allows us to tailor it enough to reduce the costs and follow better the needs that we have from the cloud market. The pace of the market is terrible; there’s massive pressure from our competitors, as well as massive pressure and load coming from our users on the new devices. OCP is the enabler,” replied  Huylebroeck.

“How do you balance the easy way of consuming a Tier 1 vs the anti-vendor lock-in story that OCP can tell?” Crawford asked Erenkrantz.

“A lot of the initial designs for OCP were really focused on the web-scale traffic, and that’s not necessarily indicative of the loads that Fin-tech companies have, and stresses in different parts of the infrastructure. Each individual Fin-tech company in the North-East wouldn’t have necessarily had the scale of Facebook. Maybe we are approaching the scale of Facebook from a volume perspective,” pitched in Erenkrantz.

“How important is to Orange the ‘lack of licensing’ in OpenStack? asked Crawford.

 .

• Licensed software will not disappear, as it’s fair paying for innovation

“It’s important to reduce that part of the cost, but I don’t believe licensed software will disappear at Orange, not entirely,” believes Jeremy Huylebroeck. “Even with OpenStack, there will be support that we get from specific players for innovation. We need to pay people to do that; it is fair. Innovation is massive in both communities. Orange really believes that the two communities should really work together, maybe more actively,” thinks Jeremy Huylebroeck.

“Is Bloomberg looking at OpenStack at all?”

“Yes, we’re having a lot of deployment in OpenStack, we’ve open-sourced all of our recipes to build our OpenStack cluster and we look at the software side, while it also makes sense for us to participate in the open hardware as well,” replied Erenkrantz.

Graham Westin, the chairman of Rackspace, described Open Compute and OpenStack as peanut-butter and jelly, noted Crawford. “Talk about rack design and how that affected your adoption of OCP,” he prompted.

“We started with the initial Facebook rack, as it would fit our business needs; open rack was almost a savior to continue to prove OCP at Orange. We are interested in the design coming up from Rackspace, that Delta is providing. We didn’t see the need in having multiple zones for our cloud services, we are a multi-tenant system that already has UPS in place in the data centers, so we don’t need batteries. In Open Rack we have the possibility to rack anything so we have solutions from almost all the vendors in OCP,” explained Huylebroeck.

“We’re using the open rack design at Bloomberg. We have a very heterogeneous computing environment where we want everything that was ever produced,” added Erenkrantz.

 .

• The need for an Open Data center movement

“One of the drags in open compute right now is these 20-year Capex investments that the data centers have made. We’ve been pushing the data center community, as they build new data centers, to build something closer to what Facebook has built, and let us come in and do the rest so that we can have these power-efficient data centers and take advantage fully of what open compute is bringing to the community. It needs to be an Open data center movement as well,” thinks Williams.

“What would be your advice to the audience in terms of being able to adopt OCP?” inquired Crawford.

 .

• Rely on the community

“If you want to adopt, rely on the community first. Don’t be afraid to join and discuss with the people. You don’t have to do everything yourself,” reassures Huylebroeck. “Rely on the vendors.”

“Focus on a particular area (storage, management),” added  Erenkrantz.

 .

• Get on board

“Riot is a very community-driven company. Being able to leverage the community with this core infrastructure that is critical to the technology that’s providing entertainment experiences, just makes sense. Aligning your business with those values throughout the entire chain, seems like a no-brainer approach. Get on board! This is going to be a big deal for your business,” promises Williams.