OVH is the number one hoster in Europe, and number three worldwide, but is it big enough to compete with Amazon Web Services, the undisputed leader in cloud computing, especially in HPC instances?

We can definitely compete on quality of service and pricing. OVH is significantly cheaper than Amazon in almost all ranges, not only in HPC. On a comparable offer, OVH is 30% more attractive. If we can beat the rates, it is because we have the critical mass, a great PUE, our own very high bandwidth network and an integrated design chain. All these aspects allow us to offer excellent service quality. We can also compete on the security part, on data location. For example, for our French customers, we can guarantee that resources and customer data will remain located in France, while Amazon has no datacenter in France. Given some recent events, it is a problem.

Regarding HPC infrastructure more specifically, we can resize at will because our cloud instances are not designed like Amazon’s. OVH gives root access to customers, so they can manage their servers as they see fit. In addition, we have on-demand jobs that include almost all numerical simulation tools. That does not exist on Amazon.

All in all, there are many points on which we are very different from Amazon. Oxalya is a pure player: our solutions were designed from the start for numerical simulation users. With Amazon, it’s rather you install your virtual machines locally and then… off you go! Of course, we also offer self-service machines, with NVIDIA-equipped accelerators and so on, but we can provide very finely tuned systems for intensive calculations or configurations specifically designed for particular uses. Amazon does not have this.

In terms of security, OVH also innovates quite uniquely. Take DDOS (Distributed Denial of Service) attacks for instance. With our infrastructure, we can take 500 Gb/s of traffic and keep the capability of sorting entry points. We know how to establish multipoint VPNs, we can provide levels of security specific to a given environment…

When Oxalya was on its own, we already used these technologies in HPC, for example to isolate InfiniBand flows. We worked hard on this InfiniBand problem, on the virtualization and network deployment automation. It is this a very specialized know-how that OVH offers globally today. What’s made by OVH and what was made by Oxalya is a perfect match.

Wasn’t there a marketing risk for a specialist such as Oxalya to be absorbed by a known heavyweight in the web hosting business?

(Smiles) The question we faced was rather “Who is OVH?”. Our customers are people who buy HPC infrastructures – generally not CIOs or web players. Besides, even among CIOs, not everyone knows OVH necessarily. When we told them what I just mentioned, the message was clear: we’ve jumped from a few hundreds to hundreds of thousands of machines. This real change in scale does not mean that Oxalya is drowned in the OVH structure. We kept all our contacts in the HPC community and therefore, on the contrary, it only strengthens and broadens our offerings. There is now a larger structure behind Oxalya but the brand continues.

For how long?

Difficult to say. At OVH, everything evolves very fast. Last year, it was decided to keep it, and a year later it is still there. We will see what happens, but the hosting and HPC markets are well differentiated, so keeping the Oxalya brand makes sense.

How did the Oxalya activity develop since the takeover? Do you have any figures in terms of customers, of HPC servers at OVH, for example?

We don’t disclose any business figures. We never communicated on this point and we do not want to do it today, all the more as we are still in launch phase. Over the past year, we focused our efforts on adapting the Oxalya software stack to the OVH platform. We also standardized Oxalya’s offers to the OVH infrastructure. That is basically the summary of our technical activity in the last 12 months. We just launched on-demand HPC “by OVH”, with OVH billing, so the current numbers would not be very significant.

But do you at least hold a significant share of the cloud HPC market in the countries where you operate ?

On-demand HPC is not a big market yet. It is still in its infancy, so speaking of market shares does not really make sense. With hindsight, I can say that when we started business in 2008, it was too soon. The situation is a bit strange, but there may be an explanation. The primary audience of HPC are people who work in innovation, academic research, industrial R&D. Paradoxically, this is a very conservative environment, at least from a technical standpoint. When I started Oxalya, my idea was that we could do HPC just like we do Web: see and work on the data online, receive a text or an email when a result is ready, etc. I was wrong. HPC users are very concerned about the location of their data, and everyone considers their data to be more strategic than others’…

Is simulation data more strategic than CRM data? For a certain number of companies, I’m not sure. The ordinary R&D culture is that everything must be partitioned and done at home. Opening is difficult, and this is where the paradox lies. At the same time, as people consume more and more online services on their smartphones and tablets, they end up thinking that eventually they could also use HPC this way. What we do today, we were able to do it five years ago. This is an environment that evolves very slowly and with great effort.

But I believe the start is taking place now. The market is ready, if only because the shifting of costs from capex to opex is interesting almost all businesses. Everyone needs to streamline and, in this context, we provide additional capabilities that are both flexible and low cost. Also, global offers have matured. Suppliers have now reached a certain size, like Oxalya with OVH. Five years ago, they were only small players. What we see today, really, is that the market is going to scale up.

Has your clients profile changed over time? Do you see more SMEs coming to HPC?

No, I do not see any profile changes for now. Today, our natural customer is a large account. But things can evolve. Originally, Oxalya was a B2B company with a very traditional approach to business: go see people, talk to them, study projects together, and then maybe get a chance to sign a contract. It was our culture. For its part, OVH is very industrialized: 90% of OVH customers order through our websites. This is something that will be new for Oxalya customers.

Where are you in your discussions with software vendors?

ISVs will be a growth driver for Oxalya and the entire OVH group. Today, we are working to implement infrastructures for big services integrators who work on complex projects. Our role is not to position ourselves on the business needs: it is the integrators, to whom we provide infrastructure, who lead the projects. So naturally we are doing the same with the ISVs that come to SaaS.

Does OVH offer the technology developed by Oxalya to “SaaSify” their applications?

Yes, even though we are positioned in different HPC areas. Our technology is not specific to simulation, we can use it with any application. Therefore, we can work with any software vendor who want their products available in Web mode in addition or in complement to classic usage. This is where we bring added value, by providing a secure, reliable and high performance infrastructure.

What simulation applications are you going to make available?

For HPC in SaaS mode, we will have stand-alone application accesses à la remote desktop. Of course, in our new series, we offer mainstream open source solutions like Code_Aster or OpenFOAM. But at the same time, we are working with a number of ISVs to allow their users to connect with us using their software tokens.

So you solved the difficult issue of usage-based pricing?

You know, to tell you the truth, some of our discussions have been going on since 2008. At the time, many ISVs would find the project “great”. But when we arrived with potential customers, most eventually found that it was “complicated” – the on-demand licenses and all that… So we decided to take another approach where we would be able to use our clients’ tokens with on-demand clusters for the requested job. Technically, there is no problem. Then ISVs have to play along. It is still too early to know what the next milestone will be. We have some quite privileged relationships with some of them, even joint projects, but as I said earlier, it takes time for things to happen. Let’s talk about it again in a few months…