The skyline query can help identify the “best” objects in a multi-attribute dataset. During the past decade, this query has received considerable attention in the database research community. Most research focused on computing the “skyline” of a dataset, or the set of “skyline objects” that are not dominated by any other object. Such algorithms are not appropriate in an online system, which should respond in real time to skyline query requests with arbitrary subsets of the attributes (also called subspaces). To guarantee real-time response, an online system should precompute the skylines for all subspaces, and look up a skyline upon query. Unfortunately, because the number of subspaces is exponential to the number of attributes, such pre computation has very expensive storage cost and update cost. We propose the Compressed SkyCube (CSC) that is much more compact, yet can still return the skyline of any subspace without consulting the base table. The CSC therefore combines the advantage of precomputation in that it can respond to queries in real time, and the advantage of no-precomputation in that it has efficient space cost and update cost. This article presents the CSC data structures, the CSC query algorithm, the CSC update algorithm, and the CSC initial computation scheme. A solution to extend to high-dimensional data is also proposed.