Artificial Intelligence

Carbon-Intelligent Computing: Addressing the demands of climate change

In April, technology giant Google announced a major paradigm shift towards decarbonized computing. They built a small, carbon-intelligent computing platform that enables servers to use low-carbon power (from renewable sources like solar and wind) for heavier tasks like filtering, searching and map updates.

The platform uses forecasts of the electrical grid’s carbon intensity from energy data specialist Tomorrow. Simultaneously, Google’s internal forecast predicts the hourly power resources a data center needs for compute tasks over the same period – allowing the system to shift tasks to minimize carbon emissions without affecting the response of the local system.

This unique approach can be implemented in data centers distributed around the globe, meaning that solar and wind energy could be generated and utilized 24x7.  Let’s look at how a program like this can be rolled out at scale.

First, it would require a similar or more optimized mathematical model to be constructed, followed by building a computing platform that uses an IoT device, as depicted in the diagram below.

Here is how it could then be scaled up:

  1. Once the model shown is deployed on one “model” server, we begin analyzing the results then make changes to make it work efficiently. Soon, we would have huge amounts of data that can be sliced based on time consumption, power consumption and frequency of process execution.
  2. We could create another model which will identify energy reserved and produced by renewable resources like solar and wind.
  3. Subsequently, we distribute power-consuming processes to another server which is running on renewables. We could also schedule these processes to run on the same server when renewable supplies are high, by switching from grid energy to a low-carbon source.
  4. We repeat these steps and fine-tune the algorithms and models until we get the desired results.
  5. We can then replicate the process across other server clusters at one site. For example, we can transfer overnight processes to a data center in another part of the world where the sun is shining — in order to use solar power to run the process.

By applying analytics, we then track and compare the carbon reductions from our energy-efficient computing model and continue to fine tune the algorithm to maximize impact.

With wind power capacity going up and the cost of generation going down, this type of dynamic power sourcing has the potential to dramatically increase the sustainability of data centers and lead to a reduction in the global carbon footprint.

It can also be applied to the manufacturing industry, which is one of the most power hungry industries and relies mostly on power generated from traditional sources like coal. The world is in the middle of a huge transition. Using green energy is the need of the hour, and companies should gradually shift from carbon-based energy sources to sustainable energy. This will help reduce our carbon footprint, become sustainable businesses and make a serious impact on the environment.