Edge/cloud resource management for time-sensitive applications

Abstract

Cloud computing has increasingly become relied upon by businesses and individuals alike for its convenience in providing resources like compute and storage. Technological advancements and skyrocketing interest in areas such as Artificial Intelligence and Internet of Things as well as the sheer increase in development of software solutions, tools, infrastructure, and more have driven demand for cloud resources to staggering heights. Serverless computing, with its auto-scaling services and pay-per-use pricing model, has also gained popularity among users looking for a hassle-free cloud experience. However, these computing models are usually associated with greater latencies due to factors such as physical distance, server load, network congestion, and bandwidth limitations. As such, time-critical applications with safety consequences provoked by the failure to meet their strict deadlines have not proliferated in the use of cloud resources. This is rapidly changing with the advent of next-generation wireless technology like 5G. For example, such applications can be found in manufacturing (real-time robotic controls) and driving automation (localization using vehicle-to-infrastructure communication).

With this gradual introduction of time-critical systems into cloud and edge computing supported by 5G wireless technology, it is crucial to ensure the adherence of strict deadlines for such systems. This project aims to develop an understanding for the network latencies experienced by a serverless application communicating via a 5G network, including latency due to network congestion and delay experienced by or priority of a packet due to the scheduling policy used.

An open source software developed for research purposes, Simu5G, is used in this project. It is a 5G NR and LTE/LTE-A user-plane simulation model and will be used to simulate and emulate a real-time 5G network.