In Cassandra, we don't get to join. This implies that we need to denormalize. In this lightning style talk, Tajinder covers the importance of designing tables around domain and access patterns. Though query first data modeling is a bit challenging, this is where Cassandra's distributed architecture shines. We are always able to scale a cluster to accommodate inevitable schema evolutions with increasing write and read traffic. It is important to learn and understand the trade-off while using materialized views and secondary indexes.