This sort of question has come up before. Here is a link to one of the times to get a sense of the position.
Note that the hardware used will have a direct impact on performance and that Drafts operates on a database rather than file scans which result in higher performance than the alternative.
How you apply meta data will also affect your options regarding how you work with the data. Constrainting by a tag and searching a subset of draft content for example should be more efficient that searching the content of all drafts.