Massively Scalable Deterministic IP Networks

Overview & Synopsis

Edge serverless computing is a promising concept in the field of cloud computing that merges the advantages of serverless computing with edge computing. This innovative approach aims to bring computational capabilities closer to the source of data or the network's edge. By doing so, it enables rapid processing in real-time with minimal delay, all while retaining the user-friendly nature and scalability that serverless computing offers. However, the research and practice of edge serverless computing is still in a relatively early stage, and the performance of current edge serverless computing platforms is still a bit far from the target. At the same time, the large-scale application of artificial intelligence, Internet of Things (IoT), and new-generation communication technologies has greatly improved network traffic but also put forward higher requirements for the response time of requests and packet loss rate and other indicators. Thus, analyzing and optimizing serverless networks at the edge becomes a concern.In this article, our main work consists of modeling and analyzing edge serverless computing networks and optimizing edge serverless computing networks using machine learning algorithms.First, we introduce the background and related technologies of edge serverless computing in detail and explain the inevitability of evolving from traditional serverless computing to edge serverless computing. After that, we propose a hierarchical structure for edge serverless computing networks based on related work and discuss the current trends in edge serverless computing and the challenges it faces.Second, we model communication and computation in edge serverless computing networks using homogeneous Poisson point process and queuing theory, respectively, and analyze the impact of different network environments on request-response events.Finally, we propose a dynamic serverless function allocation algorithm utilizing deep deterministic policy gradients and compare it with traditional algorithms.