At Cycle Computing, we know that access to compute resources at the right scale is a key driver of scientific research. Sometimes it accelerates research, but sometimes it enables work that would otherwise be impossible. A great example of this is a project recently undertaken by the NASA Center for Climate Simulation (NCCS) in partnership with Amazon Web Services (AWS) and Intel.
Researchers at NASA were studying the carbon storage in the Sub-Sahara region of Africa. The traditional method is to count and measure the shrubs and trees in the region, but over a 10 million square mile area, that’s an impossibly labor-intensive task. Instead, the researchers performed digital analysis of high-resolution satellite data.
This project was made possible by leveraging several CycleCloud features. First, the data transfer tools uploaded the image data from the NASA datacenter into S3. As file transfers completed, a custom plugin submitted the computation jobs. CycleCloud autoscaled spot instances across multiple types and availability zones in order to quickly and cheaply get the resources needed to process the data.
Once the data was uploaded to S3, scientists could re-process as needed. NCCS recently reprocessed 43 terabytes of image data in just three days for less than $2,000. This summer, the NCCS is re-processing with an additional data layer that will improve accuracy three-fold. The features and flexibility of CycleCloud will be a key part of performing this important climate research.