HOLAP storage is a combination of MOLAP and ROLAP. Detail data is kept in the relational data source. But aggregate data and indexes are loaded into MOLAP storage during processing, as shown in figure below.
When SSAS processes an MDX query, it determines where the data necessary to resolve the query resides. If it needs detail data, it translates the query to SQL and sends the translated statement to the data source. Otherwise, it retrieves aggregated data from the MOLAP Storage.
HOLAP queries are potentially slower to resolve, furthermore, storage requirements on the SSAS server are not truly lower because SSAS must read all the data from the data source to build the aggregations and indexes and requires enough space to do this. SSAS disposes of the data afterwards, thus reducing storage requirements. Because you must factor in enough storage overhead for processing, HOLAP is not a good choice if your goal is to minimize storage.