Abstract
Various types of cache eviction policies can be utilized to identify cache objects for eviction from a cache. Effective cache eviction policies that improve cache hit rates can improve performance. This disclosure describes the use of machine learning for cache management. A base policy, e.g., least recently used (LRU) or other policy, is used to organize cache objects by eviction order. A machine learning based ranking model re-ranks the eviction order such that a certain cache object is the first object scheduled to be evicted. Upon eviction, future data requests are observed and serve as feedback to improve the ranking model. The ranking model takes access patterns of cache objects to determine eviction rank.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Altınbüken, Deniz; Chen, Kevin; Gummadi, Ramki; Ju, Xiao; Sarda, Nikhil; and Song, Zhenyu, "Cache Management Using Machine Learning", Technical Disclosure Commons, (February 25, 2022)
https://www.tdcommons.org/dpubs_series/4918