[pingles]

Difficult to draw much of/any conclusion without the Redshift schema.

From direct experience of working on a moderate Redshift cluster performance is horrific once you start querying tables outside of their designed with sort and distribution keys. This is fine when you have defined use but when you're doing ad-hoc analysis or want to query across lots of dimensions we couldn't make Redshift work.

In contrast, I've seen BigQuery perform much more consistently in the face of large datasets with varying workloads.