For processing a remote database, Tableau consolidates the components in a SQL or MDX request on your VISUAL canvas. Because the database is used with stronger hardware than the laptop/working stations of the analyst, you should expect to handle queries much faster than most END users hardware-limited memory BI apps. 

For larger datasets that can be on a QUICK cluster and too big to add memory, Tableau is progressively able to shift compute (quests) close to information. Another performance factor concerns information transfer or transfer in the case of Tableau. As tableau viewing is intended for human consumption, it is DESIGNED to adapt it to the human perception system's capabilities and limitations. 

This implies usually, that the quantity of information in the query results set is low in relation to the size of the information underlying the aggregation and filtering focused on the visualization of trends and outlines. The tiny result sets require little bandwidth in the network, so Tableau can rapidly get the results. And Tableau is caching query outcomes for quick reuse, as Ross has stated. 

The final factor as stated in Eriglen includes the capacity of Tableau to speed up memory if required (e.g. when operating with very slow databases, TEXT files). Tableau's Data Engine uses memory-mapped I / O, so it can easily work with large data sets that don't fit into your memory when used in memory acceleration. Only in a subset of disk data that are needed for a given query will the Data Engine work, and as required the data subsets are mapped into memory