Case Studies

HPC – Energy Optimization

Technologies

Advanced Control System
AI & Prediction
HMI & Cloud Implementation

Client’s Need

Implement a model capable of predicting HPC power consumption by having available data on CPU and GPU temperatures and their loads. Green coding: include correlation between codes executed on HPC and energy consumption in the model. Development and implementation of a shutdown optimization policy of the idle nodes, depending on the extent of the code to be executed.
Reduction of carbon footprint.

Approach

Data acquisition campaign, data analytics Concept and AI algorithm development to estimate energy consumption of a single data centre rack, AI algorithm to optimize switching between racks to minimize energy consumption basing on job schedule of the data centre. Implemented on custom cloud (Da Vinci 1).

Main Outcomes

  • Development of a DVS ® model for consumption prediction by considering the extent of the code executed as well.
  • Evaluation of the performance of the model on test data.
  • Energy loss reduction thanks to the optimizer.