Energy is swiftly becoming a gating issue in large scale storage systems, from high-performance computing (HPC) to data intensive applications. For example, the Square Kilometre Array (SKA) [3] is a large radio telescope array expected to be finished by 2024. Its dishes will produce about one exabyte (EB) of raw image data per day. However, the power envelope goal for the storage systems of future exascale supercomputers has been set at ≈ 2 megawatts (MW) [2]. Based on this projection, we can predict that, by 2020, under the energy cap, a supercomputer will at most be able to provide 2 to 3 EB storage for the scratch space, which is far from enough for most of the planned applications. Thus, new technologies must be developed to reduce the energy consumption of future storage systems. The amount of power a large-scale computer is allowed to use is not the only concern, however. Computation has become relatively cheap, and new trends [1] suggest that data movement will be a major proportion of the power dissipation in future systems.