Content-aware prefetching in Over-The-Top wireless networks
Content-aware prefetching in Over-The-Top wireless networks, Proc IEEE International Symp. on Computers and Communications - ISCC, Heraklion, Greece, Vol. 978-1-5386-1629-1, pp. 1 - 6, September, 2017.
Digital Object Identifier: 10.1109/ISCC.2017.8024580
The Internet is increasingly considered an essential and integral part of people's lives. With its massification and the constant increase of the bandwidth offered to the consumers, services like Over-The-Top (OTT) have grown in recent years - e.g. YouTube and Netflix. OTT services refer to the delivery of audio, video and other data over the Internet without network operators' control. This trend leads to Quality-of-Experience (QoE) challenges in the presence of large delays or quality breaks in the visualization of a video stream. To overcome these issues, new approaches are needed to improve the content delivery reception to the users, and consequently, the services' QoE. This paper proposes a novel prefetching mechanism aiming to reduce the content delivery response, bringing the content as close as possible to consumers before being explicitly requested. This approach is based on the distribution of OTT contents in wireless networks, that can be integrated in network equipment: it is able to predict the type and quality of content that consumers connected to - or nearby - an access point may request, and cache it before being requested, thereby improving their quality perception of the service. Given the lack of information in the literature on management and control of proxy caches for embedded systems, the first step is to analyze and compare different proxy cache solutions - Nginx and Squid - and then evaluate two developed prefetching algorithms in mobile scenarios, given the characteristics of wireless networks. The results show that the proposed content-aware prefetching mechanism is capable of boosting the network's performance, from a technical perspective - higher hit-ratios, reduced data transfers and request latency - and from a user perspective, with significant improvements in the Mean Opinion Score (MOS) QoE metric.