Reinforcement-Learning Optimization of Storage System Configuration
A reinforcement-learning framework that dynamically tunes an operating system's storage parameters, such as cache size, queue depth, and readahead settings, by learning from real-time I/O patterns. Using deep Q-learning inside the storage kernel with low CPU and memory overhead, it replaces static heuristics with a policy that adapts to changing workloads. Students learn how storage tuning can be cast as a sequential decision problem and solved online to improve throughput and latency.
2501.00068
The exponential growth of data-intensive applications has placed unprecedented demands on modern storage systems, necessitating dynamic and efficient optimization strategies. Traditional heuristics e…