Architecture for Adaptive Live-video Streaming in Offline Dew Computing Contexts

Abstract

Live video streaming processing has become one of the main multimedia trends. The stringent requirements for low latency have led to a high dependence on cloud services, which constrains contexts where access is not stable. Thus, paradigms such as Dew Computing emerge as a potential alternative. This model proposes equipping devices with functionalities that collaborate with cloud services but become independent when the Internet is disconnected. This way, dew computing servers may process video streaming locally when the Internet is not available, but often face node mobility, which may lead to periods of disconnection. Considering these constraints, this paper introduces a software architecture to adapt live video streaming in offline dew computing contexts. For this purpose, a machine learning classification model is applied to predict link stability to adapt streaming compression accordingly. As a result, a decision forest algorithm has been applied, providing 95.9% accuracy. The obtained results are encouraging for researching local video transmission in dew computing.