From what I can tell based on research/some exposure, modeling
still happens but tends to be more abstracted than usual.
Generally the three touchpoints I see are:
-- Canonical modeling of what the data represents and how it
works across platforms to keep it relatively
-- Modeling for NoSQL databases in the mix that is direct to
schema development, and remodeling as the usage of the data derives
what the most efficient physical model is as the common usage
-- Traditional modeling downstream of the main pipelines as the
cleaner data still needs to be pulled into a data warehouse or
similar analytics platform.
The biggest takeaway I got from EDW is that modeling is still
important, and needs to adjust to gain understanding of the entire
architecture, but also get used to being agile and knowing that the
model will change frequently over time.