Enabling Data-intensive Workflows in Heterogeneous Edge-cloud Networks

The proliferation of mobile computing and Internet-of-Things (IoT) have changed devices at the network edge from terminals consuming data produced in the cloud to voluminous producers of data. Such a tremendous data growth within the proximity of end users gives birth to diverse data-intensive workflows that combine computing, analysis, and learning. These workflows challenge the conventional "transmitting data to the cloud"service mode, due to drawbacks such as large delays, bandwidth bottlenecks, and privacy issues. As a promising complement to address such challenges, the "near data processing"paradigm is quickly emerging. Through efficient mobile and edge computing techniques, IoT devices, edge servers, and cloud/HPC systems are now organically interconnected, and each data-intensive workflow could run at the most appropriate location for the highest profits. Moreover, the fusion of Artificial Intelligence (AI) and mobile/edge devices also presents many novel application scenarios and fuel the continuous booming of AI. Despite great potential, developing urgently needed dataintensive workflows such as AI pipelines in edge-cloud environments brings new challenges in service reliability and efficiency due to the inherent natures of the edge-cloud network, e.g., resource/energy limitation, device/network heterogeneity, and computing/data geo-distribution. Therefore, our research goal is to enhance data communication and processing in edge and cloud computing for data-intensive workflows and overcome the inherent reliability and efficiency deficits. To realize this goal, we integrate multiple frontier techniques, e.g., online algorithm design, network function virtualization, serverless computing, and edge intelligence (EI), towards algorithmic, systematic, and architectural achievements. Our research could be generally summarized into two complement projects. The first project is a comprehensive service function chain (SFC) framework enhancing the reliability of data communication in edge and cloud computing. The second project is a practical serverless edge computing system, which exploits new resources and architectures to enable high-quality data-intensive applications at the network edge. © 2023 Copyright is held by the owner/author(s).