Optimizing a data center is challenging. A 100 percent uptime is expected, while operational costs must be as low as possible. The balance between redundancy and efficiency is essential for having the most reliable and optimal data center.
Coolgradient helps to optimize data centers and bring them into a more ‘optimal state’ using advanced tooling like a simulation environment. Just as a pilot is assisted by an airplane’s autopilot while flying, this simulation environment helps operators run their data centers efficiently.
Optimizing a data center to reduce the use of energy requires excellent tools. Or better yet: a simulation environment that enables operators, engineers and analysts to experiment without changing anything (yet) in the actual data center. And provides the optimal settings.
Coolgradient has developed a platform which:
- Provides a simulation environment to try out settings that can reduce the energy consumption of your cooling infrastructure.
- Recommends the optimal settings under which the data center operates safely while consuming as little energy as needed.
- Increases operational stability by predicting whether safety constraints will be violated at current settings.
Autopilot for data centers
Once, flying was hazardous, but technical innovations have made airplanes the safest means of transportation. One of the most critical contributions that boosted safety and efficiency in the aviation industry was the invention of the autopilot. Today, autopilot is generally used during more than 90 percent of the flight, allowing human pilots to focus better on other operational challenges. Similarly, data center operators face various operational challenges due to the complex nature of the system. In short: now is the time for an autopilot for data centers.
At Coolgradient, we have been developing an optimization assistant to increase reliability and decrease energy usage. A system that balances out the cooling energy demand (to remove heat) with the cooling energy generated, optimizing the amount of electrical energy the cooling equipment uses.
Advanced machine learning model
At the core of this solution lie advanced machine learning models that simultaneously predict facility load and cooling energy based on various input features on cooling equipment, energy consumption and weather conditions. They act as a simulation environment that enables operators to gain insight into the effect of a change in setting without experimenting. This can be extremely useful in mission-critical environments where no risks are allowed during operation.
Using this simulation environment for one of our clients, we’ve created an Artificial Intelligence agent that experiments fór them and provides only the optimal setting changes. This agent follows a safety-first policy and always meets safety constraints because reliability is essential. It also proposes improved settings when it predicts safety constraints will be violated at current settings, thereby increasing reliability. When reliability is guaranteed, the optimization assistant finds the optimal settings to use as little energy as possible.
The result? A win-win situation with a more reliable cooling infrastructure and daily energy savings of up to almost 30%.