Using the cloud with high frequency and irregular processing.

We have to perform a lot of calculations in the market banks because we have strong needs from our operators to cover their risks, and therefore to have a certain number of prices and metrics produced on a very regular basis.
We also have regulatory obligations to explain our figures and to produce a certain number of metrics for the regulator.
So we have both a very regular production of these figures and recurring needs that are not permanent. So there are two types of needs that we have to juggle to meet the IT needs of our users.

We can therefore have different solutions depending on these two needs: we will have permanent needs with permanently available things. An internal architecture seems more appropriate to meet this need. And for a logic of irregular treatments or with frequencies higher than daily. At this point, we may have to use cloud solutions to reduce our infrastructure costs.

We have worked with several cloud players, including GCP and AWS, but we indeed liked the direction taken by Qarnot for several reasons:

• on the one hand, because of their green marker which was well put forward with their heat reuse logic,
• secondly, because it is a French player and one of our selection criteria was to choose French or European operators in terms of regulatory issues and our internal dependence, ,
• And lastly, Qarnot had a relatively interesting offer in terms of costs.

The price calculations that we will have to perform are relatively long, so our problem is mainly the calculation time on the grids. We are involved in regulatory processing and we know that we have an SLA of several hours to cover.

We have worked on our copy on several axes to have an added value to bring to our business internally to reduce costs but also for the structure in terms of energy footprint so as to respond to aspects on which our company is committed today. We can do something that is ecologically responsible while meeting a certain number of other criteria and drivers.

We are considering going more and more towards this type of solution simply because it is finally in the air of time. We have more and more needs that will come up. We're going to need the ability to capture computing power a little bit on the fly according to different needs (doing tests on a more regular basis, producing figures at the request of the regulator, etc.) and we're going to be more and more inclined towards this type of so-called cloud infrastructure.

Now we have a challenge on our side to work more and more on the optimization that we can make of these uses.