A Novel Hot-cold Data Identification Mechanism Based on Multidimensional Data
Yuanfeng Song, Shuhuan Fan, Jiaxin Xu, Jianming Liao · 2022 5th International Conference on Data Science and Information Technology (DSIT) · 2022
The effective identification of hot and cold data is crucial to improving system access performance. To address the problem of hot and cold data identification in storage systems, the limitations of the cache replacement mechanism in data storage are studied, and a hot and cold data identification mechanism based on multidimensional data is proposed. The mechanism quantifies the hot and cold data based on three features-access time, access frequency, and data dependency-to achieve the identification of hot and cold data. Experiments show that this mechanism has better recognition accuracy than LFU and LRU algorithms, and the read and write performance is also improved in most cases.