Deploying and Assessing Dynamic Tile Servers on AWS for Global Imagery Browse Services

The NASA Global Imagery Browse Services (GIBS) maintains an enormous archive of Earth observation imagery that is publicly accessible via common web services. To enable very fast access speeds, GIBS currently pre-processes all of its imagery into tiles. However, this requires a significant amount of storage for data that may be infrequently accessed. It also requires a substantial amount of effort to process and maintain that data. In addition, users lose the ability to modify data min/max scale values, color palettes, and/or band selections when the data granule is pre-processed into lower-precision image formats such as PNG or JPEG. The recent transition of GIBS to AWS infrastructure has opened up new possibilities for leveraging server technologies that utilize dynamic (i.e., on demand) tile generation as potential solutions for GIBS. We deployed a sample of dynamic tile servers into our AWS environment to ascertain their capabilities, compare performance times, and to prototype potential system architectures. Results of our investigation and recommendations for an operational system, including best practices for using cloud-optimized formats and improving performance with tile caching, will be presented.