4 of 4 people found this helpful
See Alan Eldridge's Designing Efficient Workbooks v9.2:
Firstly, the Tableau 9 data engine will run queries faster by using multiple cores where possible. The data engine can now run aggregations in parallel, splitting the work across multiple cores. By default, the maximum degree of parallelism is (number of available logical processors) / 2. This means that query operations on data extracts can run up to N times faster in Tableau 9 (where N is the number of cores in the machine).
I may have found more information in this doc 'designing-efficient-workbooks-v10.pdf'
Query elimination – running less queries
You can also see in the example above that we only executed two queries instead of three. By batching the queries together, Tableau can eliminate redundant queries. Tableau’s query optimiser will sort the queries to run the most complex queries first in the hope that subsequent queries can be serviced from the result cache. In the example, because the timeline includes Product Category and because the SUM aggregation of Sales Amount is fully additive, the data for the Category chart can be resolved from the query cache of the Timeline worksheet and doesn’t require a hit on the data source.
The query optimiser will also look for queries that are at the same level of detail (i.e. they are specified by the same set of dimensions) and will collapse them into a single query that returns all requested measures. Consider the following dashboard:
No, "parallel aggregation" is a Tableau Data Engine's feature.
What you a referring above we call internally "query fusion" - something we introduced alongside with parallel query execution and can be applied to any data source type.