Auto-scaling for a Streaming Architecture with Fuzzy Deep Reinforcement Learning

A streaming architecture is aiming to transport, process and store data and acts on real-time or nearly real-time for Big Data analytics and Internet of Things (IoT). The main requirement for such architecture to achieve its aim is the elasticity. Cloud computing is an excellent solution to satisfy the elasticity requirement. Its auto-scaling processes are allowing to automatically acquire or release resources according to the arriving workload. However, the fluctuation in scaling up and down resources is still not fully solved. We propose a novel approach called Fuzzy Deep Reinforcement Learning to scale the resources effectively and efficiently. The experimental results show that our proposed approach outperforms the existing approach based on Fuzzy Q-Learning.