I continue my side trip from my 5-part series through a second short stop. Here, I test a schema-optimized TPC-DS run on a 30-node dc2.8xlarge configuration -- the same configuration GigaOm specified in their original study.
This DC2 configuration shows the same 5.5x performance improvement when compared with the original GigaOm results, demonstrating the importance of schema and data ordering optimizations for query performance, and shedding at least a little light on the tradeoffs and use cases that could drive selection of one configuration over the other.