The lines of an internal table always have the same structure. By using extract datasets, you can handle groups of data with different structure and get statistical figures from the grouped data.
You have to define the structure of the internal table at the begining. You need not define the structure of the extract dataset.
In contrast to internal tables, the system partly compresses extract datasets when storing them. This reduces the storage space required.
Internal tables require special work area for interface whereas extract datasets do not need a special work area for interface.