Geospatial Processing in the Cloud

Cloud computing is an ubiquitous term that encompasses on-demand computing and related services over the Internet. This thesis work aims to develop a parallel program for processing geographical raster data and use it on the Amazon Web Services cloud platform for parallel, distributed processing. We will also investigate the program’s performance on the cloud platform and investigate different instances to see how they affect the program performance and network capabilities. We found that Amazon Web Services could be used for parallel processing using Message Passing Interface and that the program scales with an increased number of workers used, but that the scaling is limited by the time it takes to transfer data between nodes in the cluser. We also saw how network performance as shown by bandwidth tests varies depending on what instance is used