Sustainability in high-performance computing (HPC) has been on everyone’s mind lately. In the most recent survey from Hyperion Research, sustainability and energy efficiency rose to the top of priority lists for many HPC datacenter procurement plans, second only to performance, even more imperative than price.
Doing more with less is the new imperative for all of us in HPC.
Hewlett Packard Enterprise is ideally positioned to help you along your journey to deliver on even the most ambitious sustainability goals.
We have been transforming all aspects of high-performance computing – from hardware, interconnect, software stack, and services to make supercomputing systems both more powerful but also more power efficient. Our efforts show results – 6 out of 10 supercomputers on the top of the Green500 list published this summer are from HPE.
We continue to evolve the way we incorporate efficient power and cooling technologies for the system and datacenter, as well as energy-aware software, machine learning (ML), artificial intelligence (AI), and other techniques to optimize energy efficiency.
Measure, so You Can Change It for the Better
You can make your systems run more sustainably using HPE Performance Cluster Manager. It collects and analyzes power and cooling metrics for your system, providing direct insight into the actual energy consumption for better analysis and planning, including the ability to measure carbon footprint.
Your system administrators can use the software to optimize the energy consumption of the supercomputing installations. It allows you to make the best use of your energy resources for ever-increasing variety of workloads without wasting energy on idle nodes, while you also have the ability to adjust your operations to changing weather conditions.
Use the insights from the power and energy profiles from the applications to make your organization more environmentally responsible and switch from residency-based billing for use of supercomputing resources to billing by the energy cost of jobs.
Open Source for Sustainability
A more sustainable future is about the whole HPC community working together to lower the carbon footprint of supercomputing.
That is why HPE is an active member of the HPC PowerStack open-source Community, working towards a common software stack for HPC system power and energy management. The community is working towards consensus on which software components should be deployed and how they should interoperate to support sustainability goals across the industry.
HPE participates in the first implementation of this initiative – the Energy Efficient HPC (EE-HPC) project.
You Can’t Be Sustainable Without AIOps
For a while now, HPE has been using AIOps – or use of AI for datacenter operations to help customers forecast anomalies on power and cooling infrastructure to make the systems more resilient, efficient and help prevent downtime.
We continue to evolve our capabilities in predictive analytics and reinforcement learning to improve efficiencies in energy usage and reduce carbon emissions of datacenter operations. Hewlett Packard Enterprise Labs is investigating real-time digital data center twins that will enable optimization of your operations – from data-center layout improvements to dynamic on-the-fly workload placement to reduce your carbon footprint.
Sustainability Post Exascale
Post-exascale computing is about switching mindset from the fastest supercomputers to the fastest workflows. To enable that, we need to build tools that will make our workflows run the best – faster and more efficiently from edge to cloud.
Just as now, we expect that each customer will likely need a unique solution to help them put their own spin on sustainability.
HPE will address these needs with a holistic power management framework which will provide “on-demand knobs” of accessible performance and energy efficiency so you can optimize your operations based on your sustainability goals and needs.
We are already working with flagship supercomputing datacenters worldwide to reach this vision – prototyping application-aware and application-optimized (using ML to profile applications) real-time dynamic power and energy management solutions for supercomputers.
Future is about “Sustainability automation” – leveraging customer-defined policies to guide holistic system or even the operations-wide optimization – dynamically adopting to changing power availability, adjusting cooling, scheduling as well as moving apps seamlessly so you can deliver on your larger goals while minimizing environmental impact of your supercomputing operations.
Trust HPE to be your strategic partner for more sustainable supercomputing. For more information visit https://www.hpe.com/us/en/newsroom/sustainability.html.
Read more…" share_counter=""]