Skip to main content

News

GARR choses open source for its R&E Federated Cloud

Pubblicato il 12 May 2017

The last OpenStack Summit 2017 in Boston was the occasion to announce the first result of the collaboration between GARR and Canonical, known for the Ubuntu Linux distribution. This collaboration led to the implementation of the new Federated Cloud platform launched in November which has been designed by GARR and dedicated to the Research and Education community in Italy.

The cloud platform is based on a federated architecture design that allows institutions to contribute and share resources over multiple cloud regions. For this solution, GARR adopted the OpenStack framework and Canonical model driven operations tools, MAAS and Juju, which allow to extensively automate the whole deployment and maintenance processes. Thus, GARR has achieved levels of efficiency more commonly seen in the large data centres of Google and AWS, while keeping an agile organisation and a relatively small technical team.

As it was recently demonstrated during the last GARR workshop 2017, with this approach our staff can now stand up a cloud remotely in 20 minutes and automate all the maintenance.

Juju also allows GARR to serve the long tail of science, allowing researchers from any field to deploy self-service cloud applications, chosen from a marketplace of open source software available within Canonical’s Juju ecosystem.

We hope that this approach will facilitate the transition to cloud computing for all GARR users who need it. "The demand for cloud services in research and education is taking momentum and Italy is no exception” commented Giuseppe Attardi, coordinator of the Distributed Computing and Storage Department “GARR is applying to cloud computing the same successful community based model used for networking, setting up a Federated cloud rather than resorting just to public clouds. It is clear though that our cloud offering had to be competitive with commercial offerings, hence it was necessary to rely on extensive use of automation on OpenStack.

Canonical’s entire approach to model driven operations was our best choice to achieve the right level of economics for run our service competitively. Juju has also provided us with an application marketplace that we can offer to our users and increases the take up of our services. We are ourselves actively contributing to the Juju application, in particular for teaching and e-learning, e.g. Moodle in the Cloud and Jupyter Notebooks. Our hope is that this will pave the way towards the establishment of a community of users-developers, capable of sharing application solutions for research and education".

Further information