In general, Big-O is used to indicate how something scales with the problem size. The most common usage is for time taken by an algorithm, but it could also measure the memory used by that algorithm vis-a-vis the problem size.