A large multi megabyte L3 cache can take tens of cycles to access because of the long wires that have to be traversed. For example, it may take 20 cycles to access a 16 MB L3 cache. Instead of organizing the 16 MB cache such that every access takes 20 cycles, we can organize the cache so that it is an array of smaller cache banks. Some of these banks may be closer to the processor core, while others may be further. This leads to non-uniform cache access (NUCA), where 2 MB of the cache may be accessible in 8 cycles, the next 2 MB in 10 cycles, and so on until the last 2 MB is accessed in 22 cycles. What new policies can you introduce to maximize performance in a NUCA cache?