The time complexity of the learning phase is a very limiting factor for many systems. If the algorithm used for instance has exponential time growth, the maximum size of the training set will be quite limited. On the other hand, whether a system with complexity uses one or two hours, is usually of less importance. For a system using iteration to achieve better and better rules, the time taken into evaluation must be the time to reach a certain level of accuracy.
When considering the memory usage of a data mining system, only the complexity matters, as for time usage. Normally a system's memory requirements will only be linearly depending on the size of the data set. Still, the memory usage could give an indication to what kind of computer system is necessary in order to handle ``normal'' amounts of data.