Conceptual

Deep Reinforcement Learning for Cloud Job Scheduling and Resource Management

A survey of how Deep Reinforcement Learning is applied to job scheduling and resource management in cloud computing. Because cloud workloads and resource availability fluctuate unpredictably, static heuristic and meta-heuristic schedulers adapt poorly; DRL learns adaptive policies from continuous environment observation. The review frames scheduling as a Markov decision process (state, action, reward design), organizes DRL algorithms by methodology (value-based, policy-gradient, actor-critic, multi-agent), compares them on metrics such as makespan, latency, cost, energy, utilization, and fairness across cloud and edge settings, and surveys emerging trends and open problems.