There might be a theme to today’s posts…
Matt Allington shows us compression in Power Pivot:
Power Pivot would end up storing a table that looks more like the black table above (rather than the blue one), keeping just the minimum amount of information it needs to rebuild the real table of data on the fly when and if required. If the black RLE table ended up taking more space than the original column of data, then there would be no benefit of RLE and the original column of data would be stored. Power Pivot may use one or more of the other compression techniques used as well as, or instead of RLE – it all depends on the specifics of the actual data.
This is a very interesting look at ways the Power Pivot team optimize data storage.