Case Study: San Francisco State University achieves 20x cost savings and overnight experiment turnaround with Amazon Web Services

San Francisco State University - Customer Case Study

San Francisco State University's Computer Science department runs FEATURE, a machine-learning project (using SVMs and parallelized grid searches) that predicts functional sites in 3D molecular structures. The project's massive, repeatable computations quickly outgrew the university's shared cluster, forcing researchers to downscale experiments or wait long periods for resources.

The team migrated FEATURE to Amazon Web Services, deploying their C/C++/Perl/Python tools on Amazon EC2 with MIT StarCluster, storing databases on EBS, and using a custom Amazon Linux AMI. As a result, experiments that once took weeks now run overnight, compute costs dropped roughly 20x (from ~$1.71 to $0.08 per ECU/hour), and researchers gain on-demand scalability and faster turnaround for larger, more ambitious studies.

San Francisco State University

Dragutin Petkovic

San Francisco State University