Dynamic multi objective task scheduling in cloud computing using reinforcement learning for energy and cost...

Area of Science:

• Cloud Computing
• Artificial Intelligence
• Optimization

Background:

• Efficient task scheduling is vital for cloud computing performance, energy efficiency, and cost management.
• Dynamic workloads in cloud environments necessitate adaptive scheduling solutions.
• Balancing performance, energy, and cost remains a key challenge in cloud resource management.

Purpose of the Study:

• To introduce a novel Reinforcement Learning-Driven Multi-Objective Task Scheduling (RL-MOTS) framework.
• To enable dynamic task allocation in cloud environments using Deep Q-Network (DQN).
• To simultaneously minimize energy consumption, reduce operational costs, and ensure Quality of Service (QoS).

Main Methods:

• Development of a Reinforcement Learning-Driven Multi-Objective Task Scheduling (RL-MOTS) framework.
• Utilization of a Deep Q-Network (DQN) for dynamic task-to-virtual machine allocation.
• Implementation of an adaptive reward function considering real-time resource utilization, deadlines, and energy metrics.

Main Results:

• RL-MOTS achieved up to 27% reduction in energy consumption.
• Demonstrated an 18% improvement in cost efficiency compared to existing methods.
• Successfully met stringent task deadline constraints under varying workload conditions.

Conclusions:

• The RL-MOTS framework offers an effective solution for multi-objective task scheduling in cloud computing.
• The framework demonstrates robust performance and adaptability in heterogeneous cloud environments.
• RL-MOTS presents a forward-looking solution for next-generation distributed computing, including hybrid cloud-edge architectures.