DRL Based Multi-objective Resource Optimization Technique in a Multi-cloud Environment

The concept of multi-cloud becomes interesting progressively to cloud users because of its high response time, flexibility, high throughput, and reliability. But at the ground level, the concept of multi-cloud creates many challenges for researchers. The request of users and the multi-cloud environment is heterogeneous now a day. To work in this kind of environment required an intelligent system. Researchers are doing well in this field to make the whole process very flexible by providing an intelligent environment. The proposed multi-objective resource optimization deep reinforcement learning (MOROT-DRL) model uses the Q-learning technique of Deep Reinforcement Learning (DRL) to allocate resources in a multi-cloud environment. It includes a service analyzer for analyzing the requests and MET(Minimum Execution Time)algorithm used for scheduling the task according to execution time and then enhanced flower pollination allocate the optimized resources for the demanded request. The comparison of the proposed model is done with simulation results of MOROT and neural network model and also implemented on GoCS real dataset of google. The proposed model gives better results when compared based on energy, CO2, and cost.