While data processing engines such as Cloud Dataflow and Hadoop offer extreme computational power, they do so by following a well-defined execution plan, often with long delays in converting new data into usable insights. For many analytics workflows, this turnaround time is critical. As an example, suppose a marketing executive needs to know the effectiveness of recent changes to a marketing campaign for a given set of regions and a given demographic. Also suppose that the size of data involved is in the order of terabytes. These answers could certainly be determined using the likes of MapReduce or Dataflow, but doing so would involve developing, testing, and validating a new pipeline. If the results prompt further questions, the entire iteration cycle must start again.
For many tasks like this, a more ad-hoc and interactive approach is ideal, and data warehouse...
