This case study describes the NASA Center for Climate Simulation’s (NCCS) use of cloud bursting to count trees and shrubs in Sub-Saharan Africa. We outline the analytic methodology involved in counting and estimating tree size from satellite imagery. Applying these methods over the entire region of interest represents almost 75 TB of data and 150,000 CPU-hours of computation, a sizable effort. We detail how the NCCS successfully completed this project by utilizing Cycle Computing’s software to facilitate the bursting of jobs from the NCCS’s on-premise HPC cloud environment to Amazon Web Services. Two Cycle Computing frameworks were used en route to a solution that enabled large-scale computation and data transfer while providing an end-user experience with seamless access to cloud compute for research. This paper was presented by Cycle Computing and NASA at the 2016 IEEE Big Data Conference.
Using Cloud Bursting to Count Trees and Shrubs in Sub-Saharan Africa