HPC Cloud-based molecular modelling
- /
- POC
- /
- HPC Cloud-based molecular modelling
Albemarle Corporation is active in the development, production, and distribution of highly engineered specialty chemicals. The company wants to improve its ability to use HPC for molecular simulation through the use of GPUs, but existing simulation codes require porting to the new architectures. The final solution involves not only porting the simulation code for a hybrid GPU system, but also developing a simple user interface for preparing models and presenting them to the HPC system.
Start date: 01/07/2013
Duration in months: 18
Problem Description
Molecular modeling is a powerful tool that provides information for the design of new chemicals and materials. However, accurate molecular modeling requires high computational power. The purpose of the PoC was to port the simulation code to thetarget computer system and development of a simple user interface for preparing the models and transmitting them to the HPC system. As use-case from the oil industry was chosen: the use of catalysts for sulfur removal from automotive fuels production.
Goals
Porting on new architecture
Challenges
The challenge is to port an existing simulation code so that it will run on a hybrid HPC platform.
Innovation results
The solution involved not only the porting of the simulation code for the target computer system, but also the development of a simple user interface to prepare the models and their submission to the HPC system.
Business impact
The case study also demonstrated that the annual costs for the use of a Cloud-based HPC system on a pay-per-use basis was approximately half that of owning and maintaining a sufficiently powerful in-house system, representing a yearly saving of €38,000.