AWS Case Study: Politecnico di Milano

Hi Antonio, briefly tell us about your business.

I am an assistant professor in the nuclear engineering department at Politecnico di Milano, the largest technical university in Italy. Politecnico di Milano ranks as the 48th best technical university in the world (according to the 2012 QS World University Rankings). I work in the Nuclear Power Plant Group, which is part of the Nuclear Engineering Section of the Department of Energy at Politecnico di Milano.

Since 2000, our research has focused primarily on next generation reactors, particularly on small and medium sized reactors (SMRs) and on Generation-IV (Lead Fast Reactors and Molten Salt Reactors) studies. Current research and development themes include thermal hydraulics and safety systems, thermal mechanics, nuclear fuel performance and neutronics, model-based simulation and control, and economics.

One of the main activities of the research group is the development of multiphysics models for the analysis and design of Generation-IV nuclear reactors, emphasizing safety-related aspects. This innovative approach involves coupled simulations that consider neutronic, thermal hydraulic and mechanical aspects together.

Why did you decide to use AWS?

We needed to perform high CPU neutronic and fluid-dynamic computations quickly and affordably. We also need access to virtual servers that are easy to use and set up by researchers who do not have a strong computer science background and knowledge of cluster architecture.

We began to investigate our options. Using a virtual server outside of Politecnico di Milano for calculations that are at the heart of our innovative research was not an easy choice. However, the reliability of Amazon Web Services (AWS) as well as the large selection of Linux images made us confident.

How have you incorporated Amazon Web Services as part of your architecture?
What services are you using and how?

We use Amazon EC2 Cluster Compute instances primarily for the heavy numerical calculations required to prepare the input for transient simulations. Currently, we use Amazon EC2 Cluster Compute Eight Extra Large Instances for neutronic and fluid-dynamic computations. High CPU neutronic and fluid-dynamic computations are continuously performed in order to validate the development of our innovative approach. Figure 1 below provides a diagram of our architecture.

Figure 1. Politecnico di Milano Infrastructure Architecture

How has AWS helped your business?

With AWS, our simulations can run three to four times faster than before we started using AWS. We can often run a simulation in the morning and have the results ready just after lunch instead of a day later. We can complete two simulations per day on average using Amazon EC2 Cluster Compute Eight Extra Large Instances. The instances can be available in as little as 10 seconds, instead of a few minutes. If we want to try a change to a process or simulation, we can just create a new Amazon Linux AMI instead of changing the operating system, upgrading software, or making adjustments. Without AWS, this would be very difficult for us to do.

AWS makes it possible for us to focus our efforts on Reactor physics instead of waiting for simulation results or trying to understand the architecture of a complex computing cluster. Furthermore, before using AWS our systems were down half a day per month. Now, thanks to AWS, we haven’t lost a simulation yet.

Do you have any future plans to incorporate other AWS solutions?

We are planning to move other activities to the AWS cloud. For example, since the March 2011 accident at the Fukushima Daiichi nuclear power plant in Japan, we are paying even more attention to safety and the investigation of new scenarios with more complex physical phenomena. We plan to perform innovative multiphysics simulations on the same machines that we use for neutronic and fluid-dynamic validation. We expect AWS to be a good ally for the achievement of our future goals.

To learn more about High Performance Computing (HPC) on AWS, please visit http://aws.amazon.com/hpc-applications/.

Added October 2, 2012