Abstract

The present disclosure relates to a system and method for optimizing memory utilization using a reinforcement-learning-driven, bidding-based page-replacement algorithm in computing environments. The method involves monitoring memory usage patterns from multiple processes through a memory management unit, calculating an affinity score for each page in a buffer cache based on access characteristics (including recency, frequency, tenure, working-set membership), page-fetch/IO cost, job-submission timestamps, and physical block identifiers, and determining page-eviction decisions via a bidding process in which pages with lower affinity scores are selected for eviction. The system comprises a buffer cache for rapid data access, an affinity calculator for assigning affinity scores, and a bidding engine that facilitates adaptive eviction decisions informed by a reinforcement-learning model that is updated via a reward signal correlated with page hits and costly page faults.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Download

Share

COinS