benchmarking - TPC-DS BenchMark on Hadoop - Why use star schema -

i trying run tpc-ds benchmark sparksql.

in document talk having star schema , number of tables.

from understanding of hadoop , better have denormalized data, , can format paraquet in compression. (use partitions parallelism)

i found document sas -> https://support.sas.com/resources/papers/data-modeling-hadoop.pdf

which talks in same term. no dataware house expert, request , me understand how model data dataware house in hadoop