A single cache replacement algorithm cannot efficiently meet the needs of multiple workloads,and existing cache replacement algorithms based on policy combinations also have low cache hit rates and cannot adapt to complex workloads.To fur-ther study efficient,universal,and adaptable cache replacement algorithms for multiple workloads,a cache replacement algorithm TSCache based on policy combination thinking is proposed.By identifying and analyzing storage block level workloads,workload characteristics are extracted,and combined with Thompson sampling algorithm and regret minimization idea in strong chemistry learning to assist in decision-making cache replacement and optimize cache replacement algorithms,Being able to adaptively adopt corresponding replacement strategies under different workload characteristics improves cache hit rate and performance.The experi-mental results show that TSCache performs better than LRU,LFU,and LeCaR cache replacement algorithms when multiple load ac-cess characteristics or cache sizes are relatively small compared to the size of the workload.
万方数据