Impact of processor temperature on HPC application performance and energy consumption

Large scale datacenters manage applications as black boxes. Most of the time, they assume that application behavior is not linked to the state of the underlying hardware. When an applications runs on a hot processors, it can be slowed down arbitrary by the processor as it tries to protect itself.

The goal of this internship is to evaluate the impact of temperature on the speed of the code, the impact of the execution of the code on temperature, and the possibility to reduce the frequency of the processor to cool down the processor at key points to cool down the processor (and thus speed up the application)

This internship will:

  • Propose synthetic benchmarks stressing processor to increase its temperature
  • Acquire profiles of applications (including hardware performance counters, performance, energy and temperature)
  • Propose a model of heat production based on monitoring
  • Propose a model of the temperature impact on performance and energy of applications
  • Propose a DVFS algorithm to optimize applications from a thermal point of view

The acquisition will be done on Grid5000 (www.grid5000.fr) using Expetator (https://gitlab.irit.fr/sepia-pub/expetator)

The project is in the context of the ENERGUMEN (https://www.irit.fr/energumen/ funded by ANR) project at IRIT (http://www.irit.fr) in Toulouse in the SEPIA team. For more information contact Georges Da Costa (georges.da-costa@irit.fr). The main research topics of the SEPIA team are energy and performance optimisation of datacenter (multi-objective scheduling). Discussions are open on the exact topic as long as it stays coherent with the activity of SEPIA team.

Expected ability of the student

  • Machine learning
  • Linux
  • System
  • Knowledge on distributed systems and HPC systems would be a plus