Programs have been empirical observed to exploit temporal and spatial locality. Hence caches, which are smaller and faster data structures i.e. requiring low time to access, can be placed near the processor to mitigate interconnection bottleneck, and bridge the gap between memory and processor speed.