Challenge 12 - Predicting heat waves in computing applications - towards greener applications

[jwagemann]

Challenge 12 - Predicting heat waves in computing applications - towards greener applications

Stream 1 - Software development for weather, climate and atmosphere

Goal

Small inefficiencies in applications incrementally accrue to introduce increases in the cost of the energy used by the system – at a time when the supercomputers already require vast quantities of power to operate and cool the systems! Until recently the majority of the focus has been on lower-level hardware and software stack energy – but this isn’t the whole picture and although the software might not directly consume the energy they impact hardware performance. – less attention has been on the techniques and tools to empower application developers to understand and optimize energy resources.

Mentors and skills

• Mentors: Thomas Geenen, Atos COE
• Skills required:
  • Being able to implement solutions in C or C++
  • Some knowledge of HPC architectures and their bottlenecks
  • Familiarity with HPC ecosystems and development stacks

Note: Challenge is funded by Copernicus. Only nationals from European Union (EU) Member States and countries associated with EU’s Space Programme (currently Iceland and Norway) are eligible to participate (see Terms and Conditions).

---

Challenge description

Why do we need a solution
Significant focus traditionally has been on developing solutions to run faster and more efficiently in terms of time to solution. Increasingly the energy efficiency of applications is becoming more important to improve overall cost to solution.
Understanding potential hot spots in the codes and profiling will be an important stage in starting to improve the energy efficiency of applications. Specialist environments integrated to the hardware counter level are available but require specific technology such as FPGAs, alternatives are a number of energy-aware run time environments and understanding the current tools available and functionality and insight they provide. Developing a framework to improve understanding of the energy consumption of applications is an increasingly important operational monitoring solution.
Important considerations over-heads in monitoring real-time applications run times – as collating information should not significantly increase time to solutions for the applications.

This challenge will provide an opportunity to work with the ECWMF and the supercomputing supplier to provide technical expertise from their application specialists to assist with designing a methodology to profile the energy usage of a workflow and outline a mechanism to provide insight into a codes energy footprint.

What could be the solution
Increasingly important with the potential advent of hybrid on-prem and cloud environments, then energy consumption and operational efficiency on commercially provided platforms impact the overall cost of the calculations.
Developing a framework/mechanism that supports researchers and developers in understanding the cost of the application and identifying energy hot-spots will provide a service for code developers to investigate more energy-efficient algorithms.