6G-SANDBOX

Energy profiling and optimisation are expected to play a crucial role in realising the 5G/6G-enabled Internet of Things (IoT), as deploying intelligence closer to the network edges ensures better response times where data are generated. Despite this, research evaluating the energy performance of such deployments on next-generation networks still needs to be conducted. In our experiments, we assessed various schedulers in the Apache Storm framework, including a round-robin scheduler, a resource-aware scheduler, and PowerStorm—a scheduler designed to balance performance and energy consumption for streaming analytics in geo-distributed edge computing scenarios. For this purpose, we purchased several IoT devices equipped with 5G toolkits, configured as user equipment (UE) connected to Berlin’s Testbed network.

The installation and configuration of the 5G toolkits were completed during the project period. Additionally, we leveraged computing resources from Berlin’s core network, provided by our partners, to support our deployment. Energy consumption of the compute components was measured using smart plugs, while utilisation metrics were extracted from Berlin's 5G/6G network and the UE devices. Our findings demonstrate that distributed processing engines like Apache Storm can operate effectively over modern mobile networks, delivering high performance while adapting to different deployment configurations. Our experiments revealed that deploying workloads on UE devices with 5G network connectivity consistently outperformed deployments using Ethernet or traditional VM-based setups. This highlights the robust QoS and superior performance capabilities of 5G radio access networks (RAN). Furthermore, our PowerStorm scheduler outperformed the traditional round-robin approach regarding tuple processing while maintaining significantly lower energy consumption compared to the Resource-Aware Scheduler. These results underscore PowerStorm's ability to achieve an optimal balance.