GPU-accelerated Video Transcoding Unit for Multi-access Edge Computing Scenarios

The exponential growth of video traffic and the outburst of novel video-based services is revealing the inadequacy of the traditional mobile network infrastructure. To respond to this and to many other demands coming from today’s society, the 5G and the Multi-access Edge Computing (MEC) initiatives are proposing novel network architectures. In this context, this paper proposes the Video Transcoding Unit (VTU) application, which, leveraging on MEC principles, brings several functionalities to the edge of networks, greatly improving User Experience with mobile terminals. The VTU can be implemented as a SW (Software)-only Virtual Network Function, or be accelerated by a Graphics Processing Unit (GPU). Specific tests are described and discussed, showing the clear superiority of the HW (Hardware)-accelerated implementation in terms of computing performance and efficiency. A possible use case is presented, in which the VTU is used in a Stadium or in large public venues during crowded events like a sporting match or concerts. The work presented in this paper was undertaken under the EU Horizon2020 Sesame