Prior to 10.5, Tableau Data Extracts (.tde files) were very efficient columnar databases that performed well with the Tableau data engine. With Tableau 10.5, Tableau introduced the Hyper data engine, which shows remarkable performance gains, especially for large datasets. Both .tde and .hyper extracts will perform faster than most traditional live database connections. This is based on several factors, as follows:

• Hyper extracts make use of a hybrid of OLTP and OLAP models, and the engine determines the optimal query. Tableau Data Extracts are columnar and also very efficient to query.
• Extracts are structured so they can be loaded quickly into memory without additional processing and moved between memory and disk storage, so the size is not limited to the amount of RAM available.
• Many calculated fields are materialized in the extract. The pre-calculated value stored in the extract can often be read faster than executing the calculation every time the query is executed. Hyper extracts extend this by potentially materializing many aggregations.

You may choose to use extracts to increase performance over traditional databases. To maximize your performance gain, consider the following actions:

• Prior to creating the extract, hide unused fields. If you have created all desired visualizations, you can click the Hide Unused Fields button on the extract dialog to hide all fields not used in any view or calculation.
• If possible, use a subset of data from the original source. For example, if you have historical data for the last 10 years, but will only need the last two years for analysis, then filter the extract by the Date field.
• Optimize an extract after creating or editing calculated fields, or deleting or hiding fields.
• Store extracts on solid state drives.