From several tests, it has been observed that no single partitioning scheme performs the best for all types of applications and computer systems. Even for a single application, the most suitable partitioning technique may depend on the input parameters and its runtime state. Hence, the characterization of partitioning behavior should be based on the tuple (partitioner, application, computer system), abbreviated as PAC. Since the (A)pplication and (C)omputer system are highly dynamic, the choice of the partitioning technique should be dynamic, rather than static.

The figure above illustrates the concept of an adaptive meta-partitioner. The adaptive meta-partitioner is fed at run-time with information about the current state of the (A)pplication and the (C)omputer system. On the basis of this information, it selects the appropriate (P)artitioning technique, hence forming a fully dynamic PAC. This can be formally expressed as P_t = f (A_t, C_t), indicating that the partitioning technique P at a given time t should be selected as a function of the state of the application A and the computer system C at that given time.