Cost-Aware Dynamic Cloud Workflow Scheduling using Self-Attention and Evolutionary Reinforcement Learning

2024-09-27

International Conference on Service Oriented Computing

AI Summary

To address the joint optimization of cost and timeliness in dynamic cloud workflow scheduling, this paper proposes an online joint decision-making framework integrating self-attention mechanisms with evolutionary reinforcement learning. The method uniquely combines self-attention modeling of task dependency structures with genetic algorithm-optimized Deep Q-Network (DQN) policy training, enabling simultaneous minimization of virtual machine rental costs and deadline violation penalties. Extensive experiments conducted in CloudSim using real-world workflow traces demonstrate that, compared to state-of-the-art baselines, the proposed approach achieves an average 18.7% reduction in scheduling cost and a 32.4% decrease in deadline violation rate. These results substantiate significant improvements in both adaptability and cost-efficiency under dynamic cloud environments.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Search and Optimization: Sampling/Simulation-based Search
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Problem

• Cloud Computing
• Virtual Machine Scheduling
• Cost Optimization

Innovation

• Self-Attention Strategy Network (SPN-CWS)
• Evolutionary Reinforcement Learning
• Cost-Aware Dynamic Multi-Workflow Scheduling (CDMWS)