Answer from cs61c-aw (wing hin kao 16500429) for Question 1
since the TLB is small than L1 caches, the cost of a fully associative mapping is not too high, and a fully associative has a lowest miss rate comparing with other associative that typically used in cache.