In order to properly redesign some legacy OLAP cubes, I need to understand the general scalability and some specific drivers of OLAP cube speed:
General: How do OLAP cubes approximately scale for rows and columns (attributes)? e.g., I would assume something like n^2 or n^3 depending on attribute numbers.
Impact of hierarchical order: How does hierarchical ordering influence the calculation, storage and response time? e.g., I would assume a day-month-year hierarchy to be way faster than considering the three as separate, independent attributes.
Special cases - empty and redundant attributes: How do empty attriubtes affect the cube calculation and usage speed? What about redundant attributes' influence? e.g., regarding the latter, I'd consider redundant to have an attribute country = USA and country code = US.
