I am modeling the Gclock-pro buffer replacement algorithm in Linux kernel.
Because there are many existing modeling works on LRU, I am considering adopting LRU model to roughly describe Gclock-pro, if they are similar.
I read some papers and documents looking for their relationship.
- LRU page replacement scheme is hard to implement.
- Second chance algorithm is a good approximation of LRU.
- Clock algorithm is one way to implement Second chance algorithm.
- GClock-pro algorithm is improved version of clock algorithm with additional hands. It approximates Low Inter-reference Recency Set (LIRS) algorithm.
- LIRS outperforms LRU.
I think there is a circular relationship between LRU and GClock-pro. Unfortunately I have no idea where to start.
Can anyone give any intuition on their differences?