Integrating Serverless and DRL for Infrastructure Management in Streaming Data Processing across Edge-Cloud Continuum

Advancements in stream data processing engines (SDPE) increasingly require the use of modern Deep Reinforcement Learning (DRL)-based management strategies and distributed computing paradigms like serverless computing to enhance efficiency and scalability across the edge-cloud continuum. This study explores the challenges associated with this integration, especially with Apache Spark. Despite their advanced capabilities, modern SDPEs still lack full maturity in terms of efficiently managing dynamic resource demands and seamlessly integrating with other technologies. To fill this gap, we propose the architecture of ISIM-SDP, acronym for Integrating Serverless and DRL for Infrastructure Management in Streaming Data Processing across edge-cloud continuum. By implementing a DRL-based approach, the system dynamically adjusts resource allocation in real-time, enhancing the flexibility and scalability of computational resources. ISIM-SDP leverages serverless frameworks to reduce operational overhead and improve system responsiveness. Experimental results demonstrate the effectiveness of ISIM-SDP in optimizing resource usage and improving the throughput and latency of stream processing tasks.