I am searching for real world examples of skewed Join - the query and the datasets - to run some experiments. Till now, I have found:
TPC-DS datasets that is widely used in the database community. It is skewed on color and months fields. I briefly went over the queries that come with the benchmark but couldn't find a Join query that brings out the skew in Join. I myself created a query with 2 joins where the second Join is skewed but it is not a very natural use-case.
This blog mentions skewed Join in CDN access logs but doesn't share the dataset for privacy reasons I assume.
Any other examples out there? Maybe some users of Spark found skewed partition sizes when they are joining two datasets.