Amazon Web Services in use by researchers
Daniel Reta, Research Associate at the Department of Chemistry, recently used this service in his research. Daniel tells us about his research and his experience of using this service:
“Shrinking magnetic memory down to the nanoscale could drastically enhance the space and energy efficiencies of large data centres, and this could be made possible by using single-molecule magnets. These are molecules with carefully engineered magnetic properties so that loss of magnetic memory is slow. Unfortunately, the best examples to date can retain memory for only seconds at liquid nitrogen temperatures, making the implementation of real-world applications unfeasible. Our research focusses on methods for understanding the processes that lead to the loss of information, in the hope of proposing clear design criteria to increase magnetic memory times in molecular systems. To model the loss of magnetic information, we must explore multiple independent processes in what effectively is an “embarrassingly parallel” computational problem.
We designed a study comparing nine similar molecules with drastically different memory effects, representing a perfect test bed to asses our methods and predictions – the downside is that this study required 47,000 quantum chemical calculations, each taking ca. 4 hours on a single core in the CSF. Luckily, this type of high throughput computing is ideal to test the recently-developed Condor cloud bursting service. With the invaluable help of the Research IT, the Condor pool was extended into the Amazon cloud (AWS) and we were able to employ ca. 1.6k cores simultaneously, resulting in the completion of the 47,000 jobs in 2 weeks; these calculations would have taken 8 months on the CSF.
This demonstrates the tremendous power of on-demand, scalable high throughput computing for quantum chemistry methods, and is an avenue we will now explore further.”
Learn more about AWS cloud services at the University: