"We don't compete with commercial solutions."

Patrik Schnellmann from SWITCH spoke to him about the chances of a Swiss Academic Compute Cloud.

Patrik Schnellmann: Given that there are such successful global players, is there space for a cloud infrastructure for the academic community in Switzerland?
Edouard Bugnion: Yes, there is a big opportunity to address research problems with such a cloud, located in Switzerland and operated locally. This cloud, optimised for academic use, would be a powerful tool for data scientists. They would use it to analyse large data sets and meet all legal obligations regarding data protection. This is particularly important in sensitive applications such as personalised medicine, which seeks to optimise medical treatment depending on the patient’s genomes.

How would such an infrastructure be built?
Virtualisation is essential. Computing, storage and networking need to be systematically virtualised, moving essential functions from hardware to software and thus creating a "software-defined data centre". Using whitebox components ensures cost-effectiveness, efficiency and scalability within each data centre location. A federated approach would provide transparency across multiple Swiss data centres. This follows the approach of commercial cloud vendors and would be ideally suited to Switzerland’s political structure as a confederation.

How do you see the role and importance of open-source software and designs?
All clouds today rely heavily on open-source. At this point, this is no longer a topic of controversy. What's more interesting is the recent push towards open-source hardware such as that promoted by the Open Compute consortium. In such an approach, cloud operators contribute hardware blueprints specially geared to cloud applications. Of course, everyone benefits further down the line from the general availability of hardware that meets these blueprints.

Can a Swiss cloud for the academic community compete with the big players?
Yes, because the academic community is not trying to compete directly with commercial solutions. Instead, the goal is to serve the specific data science needs of the academic community. One specific differentiated goal is to make available interesting and relevant data sets that can be shared within the community while ensuring the appropriate level of data protection.