A Reinforcement Learning Approach for Business Processes Tasks Allocation in the Cloud

Companies increasingly rely on cloud resources for competitiveness and cost savings, but security concerns hinder full adoption. To mitigate risks, they fragment processes across multiple cloud platforms, interconnected through synchronization messages and reinforced by separation constraints. Exploring machine learning techniques, particularly Reinforcement Learning (RL), becomes essential for addressing optimization challenges in dynamic cloud environments, handling data flow, scalability, and security. This paper proposes leveraging Q-learning, a form of RL, for efficient task allocation in cloud computing. Unlike traditional methods, this approach ensures effective task assignment management in complex and evolving environments, reducing information overload per cloud. Results demonstrate the superiority of our proposed model over constraint-based approaches in generating Business Process (BP) fragments with higher obfuscation and reduced execution times.