An area to discuss use of ClusterODM
I’m new to ODM and am running on AWS EC2 instances against engineering survey data (621 images in my test set). I’ve run 3 variants: large compute optimized, multi-tasked mid-range memory-optimized, and cluster (parallel) using identical small end memory optimized. I’m running under AWS Linux 2. The only very minor hitch for clusterODM was that the AWS Linux required a telnet/xinetd install and service start. I’m creating images (AMI’s) as I go so that I can launch large clusters of machines easily. I plan on using the infrastructure as code approach with yaml eventually. My original track on image processing was following the Spark work being done within OSGeo. I am curious about what people think the performance differences will be between a Spark-based approach to parallel image processing where images are mapped to data frames (Raster frames) and the split-merge approach in ODM. Of course, image processing covers a lot of territory but would be interesting to hear where one is better than another. I look forward to participating with a great community like this.