Abstract: A data cube of n dimensions has 2 n views The more the views, the more the maintenance time of data cube By dividing dimensions of a data cube into partition dimensions and no partition dimensions, a data cube can be transformed to a star cube, which includes a summary table and views consisting of partition dimensions Star cube uses prefix and tuple sharing technology to reduce the storage requirement, computation time, and maintenance time By confining the size of a fragment to one I/O unit, the query response time of star cube is almost the same as data cube The results of experiments also show that star cube is a promising way to implement data cube