Relevant quote: "LRU does worse than 2-random when the miss rates are high, and better when the miss rates are low."
"The miss rates are high" is the case when evicting newer but unused in nea future objects is preferable, in my opinion. Thus, "two random choices" algorithm is a direct competitor to SIEVE, they work better than LRU in about same conditions.
Here are some comparisons of 2-random-choices with LRU in context of CPU caches: https://danluu.com/2choices-eviction/
Relevant quote: "LRU does worse than 2-random when the miss rates are high, and better when the miss rates are low."
"The miss rates are high" is the case when evicting newer but unused in nea future objects is preferable, in my opinion. Thus, "two random choices" algorithm is a direct competitor to SIEVE, they work better than LRU in about same conditions.