The Power of Cloud Computing
We at ChemAlive recently launched our first software module called ConstruQt. It is a high throughput quantum chemistry software that scales on the cloud. When Laetitia and I began building this software in 2015,we used a 16 core Dell Server in L’s living room. In the summer she would complain it was too hot. Our early version produced a database containing about 130,000 molecules and could only add 10,000 at a time due to storage restraints. This took over a year to accomplish.
At last count, ConstruQt has produced a dataset of 3.5 million molecules, 200,000,000 structures and is approaching a terabyte in size. As far as we know, this is the largest dataset of quantum chemical data out there. We accomplished this in its entirety in 2 months at the end of 2018 on an AWS Spot computing cluster supported by an AWS Activate Grant Scheme.
Piz Daint is the largest supercomputer in Europe located in Switzerland (named after a mountain peak, obviously). Imagine you are an academic who needs computing power. You must apply for Piz Daint time and share its resources with thousands of other users. It will take months to have your application accepted/reviewed and the system architecture will be controlled externally. For example, with arbitrary wallclock times for processes (about 6 hours, I hear).
With AWS we achieve 10% of this computing power of Piz Daint with a wait time of about 2 minutes to warm up – and its all ours. This is accomplished with our specialized Spark cluster running our core code on 8,000 cores. This distribution allows us to process molecules with accurate mathematics and modern quantum physics at the same order of magnitude as using classical Newtonian physics – currently the go-to methodology in drug discovery despite known inaccuracies and scope limitations compared to quantum chemistry.
The cloud revolutionizes access to computing power and makes our company competitive and state-of-the-art with a few clicks on a web app.