Renewable-Aware Edge Computing

Future application domains such as the Internet of Things (IoT) demand highly distributed data processing close to the edge of the network. At the same time, energy generation is becoming more and more distributed too, due to the adoption of renewable sources such as roof installations of photovoltaic panels. Goal of this research is to use co-simulation to identify and exploit synergies in these two areas in order to reduce their carbon footprint and increase resilience of critical edge infrastructures.

Motivation

The adoption of renewable energy sources such as solar and wind is a key factor in mitigating the effects of today’s climate crisis. However, power grid operators are struggling with this fundamental change in energy supply since renewables are inherently more distributed and variable than conventional sources like nuclear, coal, or gas. Maintaining a stable power grid is becoming increasingly difficult since classical consumers such as households can suddenly turn into prosumers that might overload the network with the energy they produce.

Simply storing all this excess energy isn’t feasible with today’s technology, which is why research and industry additionally focus on dynamically managing the demand-side of the power grid. Contracts are already in place that allow electricity suppliers to reduce the amount of energy provided to major consumers such as steel factories, refrigerated warehouses, and data centers when the grid is overloaded. While most demand-side management efforts are sensibly focused on these large energy users it is expected that also smaller consumers will be controllable in the future due to the adoption of connected devices in the context of smart grids. For example, more and more households are equipped with smart metering and smart devices that can be remotely controlled to react to the state of the power grid.

Besides energy supply, also computing is expected to become more distributed and heterogeneous in the future. The increasing digitalization of urban and industrialized environments demands less centralized ways of processing data, since sending large volumes of raw sensor data directly to cloud providers causes latency and bandwidth issues and raises privacy concerns. Synergies between these emerging paradigms, often referred to as fog or edge computing, on the one side and distributed power generation on the other side, suggest themselves but are barely researched.

Approach

By shifting computational load towards times and locations where there is a lot of green energy, we cannot only stabilize the power grid but also increase the usage of renewable energy sources, thereby reducing the carbon footprint of distributed computing systems. Furthermore, renewable energy sources like photovoltaic (PV), in combination with energy storage, can help to operate IoT and edge computing infrastructures that have no or no continuous access to the power grid. Prominent examples of such energy-autonomous systems are connected vehicles.

We want to identify use cases and develop concrete architectures and algorithms that allow the emerging areas of IoT and fog/edge computing to be closely aligned with and benefit from distributed power generation. Rsearch questions include:

Results

To facilitate research on energy-aware computing we introduced LEAF, a simulator for modeling large energy-aware cloud, fog, and edge computing environments. LEAF features a holistic but granular energy consumption model that covers data centers, edge devices, and network as well as applications which are running on this infrastructure. By combining analytical and discrete-event modeling, the proposed model enables the simulation of thousands of streaming applications on a distributed, heterogeneous, and dynamic infrastructure. The publicly available implementation of LEAF was evaluated within a realistic smart city traffic scenario and proved to be a valuable tool for research on energy-conserving fog computing architectures, task placement strategies, and energy-saving mechanisms.

Publications

Contact

If you have any questions or are interested in collaborating with us on this topic, please get in touch with Philipp!

Acknowledgments

This work was supported by the German Academic Exchange Service (DAAD) as ide3a.