Datacenter Cooling System

#Cloud
#IoT
#Machine learning

Business Challenge

Heating, ventilation, and air conditioning (HVAC) are one of the most critical systems in a data center. But while it protects the equipment that works under load, it’s also extremely expensive in operation. Continuously changing datacenter configurations, loads and environments prompt us to develop new HVAC control solutions.

Modern data centers need smart microclimate control systems to prevent data loss or damage caused by overheating. Designing and developing a system like that is challenging but rewarding, and this is exactly what we set out to do for one of our clients. Together, we built a more environmentally-friendly solution that meets SLA requirements and cuts down cooling unit loads and huge electricity bills.

Solution Overview

The solution we’ve developed relies on the concept of proactive control instead of a reactive one. We used IoT and machine learning to implement it.

The IoT system collects data from the sensors and the dynamic power consumption from the power supply device.

A pre-trained ML model predicts temperature changes based on sensor data and defines the optimal rack fan speed.

The solution also gives operators an interactive dashboard with all sensor data and an automatic/manual control mode.

Let's discuss your idea!

Technological Details

We applied the following tools and technologies to develop the system:

The control panel
We used Python as the main language for the server part, and we also provided the Machine Learning part with Python tools. Eclipse Mosquitto was used as a data broker. For data visualization, we chose Tornado and Bokeh.
The IoT subsystem

The hardware includes thermal and humidity sensors that are connected to a Raspberry Pi microcomputer as well as a UPS. Raspberry Pi controls rack fan speed.

We used Adafruit to get data from sensors and MQTT Paho to send it to the server for future analysis.