The past decades have witnessed the dramatic increase of smart mobile devices (e.g., smartphones, tablets) requesting multimedia services that generate the bulk of data traffic. Because of the revolution of multimedia services (e.g., on-demand video streaming, social networking, and virtual reality), the focus of network design has shifted from making a connection to delivering content. Cache-aided content distribution network architectures are emerging as an innovative solution able to harness device memory, a cheap and widely available resource, into bandwidth, so as to meet the dramatic increase of data traffic.
In the past years, numerous cache-aided network architectures have been proposed. These caching approaches usually consist of two phases: placement and delivery. The former is executed during off-peak hours, when popular content is duplicated and stored in the mobile devices. The latter usually occurs during peak-traffic hours, when the actual users’ requests are revealed. In particular, if the requested content is available in the users’ neighbourhood, it can be directly served via local links. As such, mobile edge caching/computing is emerging as a key enabler to bring the content closer to end user devices. This has a two-fold advantage: 1) the backhaul network throughput is reduced, thus alleviating the network congestion during peak-traffic hours, and 2) the radio spectrum can be very densely reused since a large number of local links can share the same bandwidth without causing significant interference.
The project aims to construct a theoretical framework for designing and developing efficient yet low-complexity combinatorial algorithms for mobile edge caching/computing in large-scale content distribution networks. One of the key enablers is the extraction of structural features from the underlying communication networks, such as content popularity, user demands patterns, and partial network connectivity, in order to jointly optimise content placement, prefetching, and delivery. In such a framework, it is promising to exploit the structural properties of network topology, user demands redundancy, and content popularity, as well as the integrated advantages of edge caching, network coded multicasting, and distributed computing. In particular, the distribution network connectivity will be modelled as a partially-connected bipartite graph, where the distributed content servers are on one side and user devices are on the other side, and a server is connected to a user device if and only if they are physically reachable one another. The user demands pattern will be modelled as another partially-connected bipartite graph, where the packetized file segments are on one side and the users are on the other side, and the connections between file segments and users represents the user-file demands.
This project will concentrate on both theoretical foundations and combinatorial algorithm design, aiming to understand the interplay between network connectivity and user demands pattern, and the roles of structural properties in determining the information-theoretically optimal trade-off between devices’ cache memory size and the responding content prefetching/delivery delay.
For academic enquires please contact Xinping Yi ([email protected]
For enquires on the application process or to find out more about the Dual programme please contact [email protected]
When applying please ensure you Quote the supervisor & project title you wish to apply for and note ‘NTHU-UoL Dual Scholarship’ when asked for details of how plan to finance your studies.