Downloading TCGA esophageal carcinoma counts, using a density heatmap to show the samples are not comparable, and then watching what CPM, upper quartile, TMM, median of ratios and quantile normalization each do to the distributions.
Range joins, high-cardinality group-bys, a 46M by 36M join, and a full analysis, each timed separately for loading and computing.
Strategies for working with large data with limited resources