V2GRAN

BASIC INFORMATION

Short name: V2GRAN
Long name: Energy-aware Autoscaling of Virtualized Radio Access Networks

Company: TSSG, Waterford Institute of Technology
Country: Ireland

Call: F4Fp-08 (see call details)
Proposal number: F4Fp-08-M26

SUMMARY REMARKS & TESTBEDS

The next generation of the 5G network will have to handle large volumes of data generated from millions of devices. To accommodate these next-generation services, operators will need to rapidly evolve their mobile network architectures to process high bandwidth real-time data. This means transforming the core telecom network from a fixed-function hardwired appliance architecture to a software-based open hybrid cloud platform. One way that operators are looking to streamline the network is by virtualizing baseband units in the radio access network called asvBBU. Introducing vBBUs is an essential step towards5G and beyond because it simplifies the deployment of new features and algorithms by streamlining resource use. The Cloud Radio Access Networks (CRAN) separates network functions from the underlying hardware, making it possible for a flexible and dynamic RAN environment. This design also leads to cost-effective networks supporting various access technologies such as millimetre waves, LTE, and Wi-Fi Essential for 5G. Reference open-source implementations for CRANs are available for researchers and developers to build highly scalable and efficient radio access networks. This project aims to implementing and testing energy-aware resource management algorithm using machine learning. Previous results show that the use of autonomous scaling and intelligent load distribution algorithms reduce energy use by more than 20% and carbon emissions by 14%. Use of machine learning-based auto-scaling reduced latency and jitter by up to 6% and 13% respectively. In this project, we will exploit additional scalability of CRANs to utilize the available resources more efficiently, thereby reducing energy consumption and carbon emission. We intend to use two testbeds for this experiment; i.e. IRIS and Planet Lab Europe (PLE). The reconfigurability of virtualised radio hardware in the IRIS testbed will allow us to implement the proposed self-orchestration mechanism for autonomous allocation of processing resources in a CRAN. We will exploit the virtual machines’ availability of the PLE (European arm) for the geographically dispersed migration of virtualized baseband units (vBBU). This will allow us to monitor the effect of migration on the QoS and to measure the change in carbon emission. As the servers in PLE are located at geographically dispersed locations and each location has different sources of energy and different value of carbon emission, therefore migrating vBBU among these servers based on the current value of carbon emission can help reducing carbon footprints. With the successful implementation of an intelligent energy-aware auto-scaling mechanism in OAI reference design, telcos and communication service providers could have an agile future-proof energy-efficient platform that they and their customers can use to accelerate innovation towards5G and beyond.

MATERIALS