Serverless Computing with Ubiquitous Cloud: Opportunities and Challenges

The rapidly emerging multi-layer hierarchical cloud and network infrastructures have enabled cloud computing to evolve from a data center-centric model to the ubiquitous cloud. Serverless computing, also known as “Function as a service (FaaS)”, inherently aligns with the requirements of ubiquitous cloud applications. However, existing FaaS programming frameworks exhibit limitations in large-scale cloud service deployment, geo-distributed resource management, and function runtime abstraction. For example, the constrained CPU resources of individual functions struggle to meet the demands of complex tasks, and there is a lack of industry practice in cross-region function scheduling with heterogeneous hardware. Moreover, inter-function communication overhead and long cold-start latency can significantly degrade the user experience. This leads to a challenging and urgent problem for cloud providers to exploit the potential of serverless computing in such scenarios. In this paper, we present several system-level analyses and discuss the design of a generic proposed cross-region FaaS framework that aims to facilitate the deployment of distributed cloud services under SLA guarantees while achieving high resource utilization on ubiquitous cloud infrastructures. Finally, we further propose a future development of intelligent ubiquitous cloud ecosystems.