Efficient Scheduling of Multiple Data Transfers in Mobile Applications

Abstract

Over the last decade wireless data streaming has progressively become an important aspect of our life. However, the lack of appropriate mechanisms to handle wireless network disruptions and delays, causes mobile applications to frequently suffer from unstable wireless connectivity, which makes data streaming particularly challenging leading to unstable streaming rates, long delays and often network failures. This paper proposes a framework that utilizes performance monitoring, machine learning and scheduling techniques to effectively schedule data streaming for mobile applications. The framework monitors the current network conditions through a set of performance metrics and evaluates its availability using a machine learning approach. To meet real-time performance objectives, if the network is characterized eligible for streaming, we use a deadline-first scheduling order for the data streaming. Our experimental evaluation, using different network traffic scenarios, illustrates the performance and benefits of the approach proposed.