Title: Video Streaming in 5G Wireless Networks
The latest work from the Network Systems Lab conducted by Saleh Almowuena, a PhD student under the supervision of Dr. Hefeeda, secured the best paper award in the ACM Multimedia Systems 2015 Conference--one of the main conferences in the multimedia research area.T
The number of mobile devices is expected to be 11.5 billion (~1.5X the World’s population) by 2019, and most of them will be multimedia capable. Moreover, most users expect improved and rich video quality as they rely more on their powerful mobile devices (e.g., iPhone 6 already supports HD video). This creates a major challenge for cellular networks as it substantially escalates the load on them. To address this challenge, Dr. Hefeeda’s research group is designing novel algorithms and systems to enable current and future cellular networks to increase their capacity, optimize the quality of video delivered to users, and extend the lifetime of the batteries in mobile devices.
In their award paper, Almowuena and Hefeeda introduced the novel idea of dynamically configuring cells to collaboratively form single frequency networks (SFNs) in order to serve large number of mobile users requesting videos. They designed an efficient algorithm to determine
the number of SFNs that should be created in the network and which cell should belong
to which SFN. In addition, the algorithm decides, in real-time, for each video request,
whether it should be served using unicast, multicast within a single cell, or multicast
across an SFN. This is a challenging problem because of the many parameters in different
layers that need to be considered to compute efficient solutions. Their results show that
orders of magnitudes increase in the service ratio can be achieved in 4G networks compared
to the current state-of-the-art.
Furthermore, the benefits of their work will further be amplified with the expected new
features in 5G wireless networks, which include massive MIMOs (antennas) and base station
virtualization. Massive MIMOs will allow each base station to have tens to hundreds of
antennas, compared to only a few in current base stations. This will introduce more
opportunities for optimizations. For example, one base station can join multiple SFNs
using different MIMOs, while at the same time it can have other MIMOs to serve local
traffic. Base station virtualization will further increase the ability of a cell network
to optimally configure itself to meet the dynamic user demands, as well as reduce
deployment and operational costs.