Public defence: Andres Felipe Ocampo Palacio
Andres Felipe Ocampo Palacio defended his thesis "On the realization of Cloud-RAN on Mobile Edge Computing for 5G and beyond generations of mobile systems".
Andres Felipe Ocampo Palacio defended his thesis "On the realization of Cloud-RAN on Mobile Edge Computing for 5G and beyond generations of mobile systems".
Trial lecture: "The role of Open Hardware and Software in the 6G Era: Challenges and Opportunities".
Ordinary opponents:
The chair of the committee: Associate Professor Boning Feng, OsloMet
Main supervisor: Haakon Bryhni
Co-supervisor: Ahmed Elmokashfi
The cellular network architecture is evolving to support a wide variety of applications with different traffic characteristics expected for 5G and beyond. Sharing computing and network resources, Cloud based Radio Access Network (Cloud-RAN) in conjunction with Mobile Edge Computing (MEC) are considered key enablers to build 5G networks in a cost-efficient way.
Understanding the limits and constraints of deploying the Cloud-RAN on MEC servers allows the system to be engineered meeting latency and capacity requirements. In this thesis, we focus on sharing computing and networking resources in MEC servers, which run a software implementation of the Base Band Unit (vBBU) along with collocated applications.
First, this thesis discusses the deployment of a small scale 4G/5G testbed based on open source vBBU and core network. In particular, we deploy a mobile system testbed which integrates the Cloud-RAN architecture with a core network deployed as virtual network functions (VNF).
While the Cloud-RAN deploys two vBBUs with different functional splits and switched Ethernet mobile network, the core network deploys VNF components using state of the art open source Management and Orchestration platforms (MANO). Setting up 4G/5G experimental infrastructures integrating these technologies, is necessary to optimally design 5G networks.
Subsequently, we evaluate the feasibility of using switched Ethernet for aggregating multiple vBBUs into the same transport network. Furthermore, we evaluate NIC sharing mechanisms allowing vBBU's time-sensitive traffic sharing network resources with collocated applications.
To run the vBBU on a MEC server, the host OS must provide RT guarantees: preemption, and a scheduling policy that focuses on meeting timing constraints of individual processes rather than maximizing the average amount of scheduled processes.
In this thesis, we assess the impact on the vBBU caused by sharing computing and networking resources in a MEC server managed by the Linux RT-Kernel. Specifically, we study the processing latency of vBBU functions which run as RT threads in the RT-Kernel. Also, we study CPU mapping of vBBU's threads as an alternative solution to mitigate the impact of sharing computing resources.