In the latest version of 6SigmaDCX, we’ve introduced a new feature that allows external systems - such as MATLAB - to read from and write to the data center model dynamically. MATLAB gives you a greater level of control over the simulation, even allowing changes to be made while it is solving! This allows for a deeper understanding of the room’s behavior with different load distributions. MATLAB also interactively enables advanced PID control systems to be input to and tuned within the model, giving controls engineers the capability to evaluate the performance of even the most complex setups.
Advanced 6SigmaRoom Modifications via MATLAB
MATLAB enables you to dynamically exchange information with the 6SigmaRoom CFD model. With this capability, you can monitor every component in the Digital Twin and alter control systems and component power dissipation - all while the model is solving. For example, you can preconfigure a change in the Digital Twin’s power profile during a transient failure scenario. Similarly, you can monitor and iterate the room’s power profile while reporting the facility’s performance, enabling rapid optimization of the room’s cooling system so that it can handle various IT power distributions.
Perhaps most critically, the communication between MATLAB and Room enables you to extract key performance metrics - such as maximum IT equipment temperature or average flow across floor grilles - from the Digital Twin. You can then send this information to an external algorithm, and use the output of that algorithm to update the Digital Twin; for example, with an updated IT load distribution within the facility. This is designed so that even newcomers to MATLAB can easily implement this type of communication with just a few lines of code.
Figure 1: The code between the green “Your Code Here” comments (top image) is all that’s needed to throttle two cabinets’ worth of IT from 100 % to 10 % heat power factor at the 50th iteration (bottom image). *Note the drop in sensor temperature of the IT associated with this action.
Test & Refine Complex Control Systems
Highly complex control systems can now be deployed effortlessly, granting limitless design capabilities to controls engineers. This not only allows you to implement advanced PID controls in 6SigmaRoom, but also allows you to verify the performance of a given controller configuration in a system with many independent variables. The response to a given controller setup enables controls engineers to extract valuable design information about the Digital Twin, then use this to fine-tune a control system using the Control Systems Toolbox available in MATLAB. As a result, a controller network design may be iterated and tested to perfection in the Digital Twin before being deployed in a live data center.
Figure 2: An example of the MATLAB Control Systems Toolbox where the user is tuning a nonlinear system in the controls GUI on the basis of system rise time .
Intelligent, Risk-Free Design & Testing
6SigmaRoom already combines unbeatable accuracy with unprecedented speed in solving the CFD model of the data center. When the highly versatile, easy-to-use optimization tools offered by MATLAB are combined with 6SigmaRoom, this grants you a new level of design freedom. Chiefly, it allows for dynamic interaction with the Digital Twin while the model is being solved, and for advanced control systems to be designed and tested in a risk-free manner. With these tools, and the ongoing development of new features to further bolster ease of use, MATLAB’s interactivity with 6SigmaRoom brings the industry a step closer to enabling AI and machine learning in data centers.
 https://www.mathworks.com/videos/nonlinear-plant-control-at-different-operating-points-68747.html from timestamp 3:18/4:22
Blog written by: John Kearns, Applications Engineer
Other Recent Posts
Unpredictability is public enemy #1 for the data center
If the pandemic has taught us anything, it’s that for all of the experience and intelligence in bus…
Power is in the firing line for risk averse data center operators
There’s never a good time for an unplanned outage. A lack of visibility or capacity to report is ne…
24 June, 2018