Many community members are using Direct Discovery to their advantage. I'm sure one of them will post his/her experiences with this method of obtaining data sets.
Keep in mind that Direct Discovery is not a replacement for the normal QlikView Script-controlled data extraction process. Direct Discovery can be used to speed up the performance of your document in very specific cases, for example when in need of a limited set of transactions from a huge data source based on specific selections in your document. In that case, loading everything beforehand into your QlikView document would make it huge (and possibly sluggish if your server isn't scaled accordingly) while only a limited number of rows is needed at any time.
If you try to use Direct Discovery to offload a substantial volume of data rows from one or more data sources, you will face typical database query performance issues, like any other database client. If the code to fill a view is long-running, users will have to wait for the view code to complete before even the first row is transferred. If Direct Discovery needs to transfer a large number of records from the same data source, the DB engine and network may become bottlenecks when multiple users start requesting rows at the same time.
Compare this with the high performance in-memory analysis that QlikView offers out-of-the-box whereby all data is present in RAM (the fastest storage medium available at the moment). Direct Discovery is not an end by itself.
Best,
Peter
