Method and system for prefetching data to cache
By dynamically adjusting the caching strategy and utilizing correlation graphs and time mutation point identification techniques, the cache queue is optimized, solving the data access problem of existing caching strategies under high dynamic loads and improving cache hit rate and resource utilization efficiency.
CN122285546APending Publication Date: 2026-06-26BEIJING INTERSTELLAR INTELLIGENT STORAGE TECHNOLOGY CO LTD
0 Cites 0 Cited by
Patent Information
- Application Number
- CN202610622580.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-05-08
- Publication Date
- 2026-06-26
AI Technical Summary
Technical Problem
Existing caching strategies cannot be dynamically adjusted according to actual request patterns, resulting in low data hit rates and high replacement costs. They are particularly difficult to adapt to data access patterns under highly dynamic and volatile workloads.
Method used
By acquiring and sorting the initial dataset, access record items are extracted, address association rules are mined, an association graph is constructed to identify time mutation points, data block priority weights are dynamically adjusted, and the cache queue is optimized.
Benefits of technology
It improves cache hit rate and resource utilization efficiency, adapts to changes in access patterns, and enhances the real-time performance and rationality of caching strategies.
✦ Generated by Eureka AI based on patent content.
Abstract
The invention relates to the technical field of data prefetching to a high-speed cache, and discloses a method and system for prefetching data to a high-speed cache, and the method comprises the steps: sorting an initial data set according to time, dividing the initial data set into sub-data sets, extracting addresses and timestamps in the sub-data sets to form access record items, mining association rules, and removing exceptions; and obtaining a final record item. Clustering and grouping are carried out according to the one-dimensional coordinates of the data blocks, correlation coefficients between core points are calculated to construct a correlation graph, and time mutation points are recognized by comparing singular value vectors of adjacent windows. And correcting the priority weight of the data block with the average access delay exceeding the threshold before and after the abrupt change point, and rearranging to obtain a final priority queue. And taking out a preset number of data blocks to form an initial cache queue, eliminating zero hit blocks through multi-round simulation, and reordering according to a hit proportion correction weight to obtain a final cache queue. According to the method, the cache strategy can be dynamically adjusted according to the actual request mode.
Need to check novelty before this filing date? Find Prior Art