The 5G-MEDIA project has recently published three articles, led by University College London, that re-imagine how advanced applications are placed in the network for improved user Quality of Experience. In 5G-MEDIA, applications are deployed as virtualised software components and embedded in edge data centres in 5G networks and throughout the Internet. The advantages of using virtualised components is that they can be deployed rapidly and can be positioned according to dynamically changing user demand patterns. This raises new opportunities for applying advanced optimisation algorithms for the placement and selection of media services to trade-off user experience, system performance and operational costs.
In “Utility-centric Networking: Balancing Transit Costs with Quality of Experience” published in IEEE/ACM Transactions on Networking, we propose a new framework that takes into account the utility of particular QoS metrics for highly demanding services, e.g. tele-immersive 3D gaming needs low delay and ultra-high definition video streaming needs high bandwidth. We propose new algorithms to optimise service placement and selection, where users are directed to the best copies of the service. Our approach also opens the possibilities of redesigning Software Defined Networking controllers to tailor routing and path selection to the QoS needs of different media applications.
In “Utilitarian Placement of Composite Services” published in IEEE Transactions on Network and Service Management, we extend this framework to composite services made up of several autonomous software components running in different parts of the network. Given the complexity of this problem we applied state-of-the-art machine learning techniques to achieve efficient service placement with low start-up delays. These techniques are being applied to ultra-high definition content distribution systems in 5G-MEDIA for dynamic cache management and for personalised content adaptation.
Finally, in “DR-Cache: Distributed Resilient Caching with Latency Guarantees” presented in IEEE INFOCOM 2018, we centre our analysis around network resilience to build service placement algorithms which minimise the probability of disruption when links fail. This work will be crucial for services that require high availability, e.g. remote production systems for broadcast-quality media, multimedia for healthcare, and traffic control systems for autonomous vehicles.