High-performance computing (HPC) is the use of specialized hardware and software to perform complex and intensive computations at high speed and scale.

HPC is essential for research and academia, as it enables scientists and scholars to explore new frontiers of knowledge, solve challenging problems, and make groundbreaking discoveries.

However, HPC also comes with high environmental and economic costs. HPC systems consume a lot of electricity and generate heat.

They need cooling and ventilation, which emit a lot of carbon and cost a lot of money. These effects worsen climate change and make HPC less accessible and affordable.

Green HPC is an emerging field that aims to address these challenges by designing, operating, and evaluating HPC systems in a more energy-efficient and environmentally friendly way.

It can reduce its electricity consumption and carbon footprint while maintaining or improving its performance and quality.

Green HPC can also lower operational costs and increase the availability and reliability of HPC, making it more accessible and affordable for research and academic purposes.

In this post, we’ll introduce the concept of green HPC and explain how it differs from traditional HPC. We will also discuss why green HPC is critical and its environmental and economic benefits.

We will then highlight some of the most promising green HPC technologies and their applications in various domains of research and academia.

Finally, we will provide some practical tips and strategies for implementing green HPC in your own projects and settings.

So, if you want to learn more about green HPC and how it can help you accelerate your discoveries, read on!

What is Green HPC?

Green HPC is a term that refers to the design, operation, and evaluation of HPC systems in a way that minimizes their environmental impact and maximizes their energy efficiency.

Green HPC systems aim to reduce the electricity consumption and carbon emissions of HPC while maintaining or improving its performance and quality.

Green HPC differs from traditional HPC in several aspects, such as:

• Design: Green HPC systems use hardware components that are more energy-efficient and environmentally friendly, such as low-power processors, liquid cooling, and renewable energy sources. Green HPC systems also use software tools that optimize the code and algorithms for energy efficiency, such as energy-aware scheduling, dynamic voltage and frequency scaling, and MLPerf.
• Operation: Green HPC systems monitor and control the energy consumption and performance of the system and adjust the parameters and settings accordingly. Green HPC systems also use system-level solutions that improve the cooling and ventilation of the system, such as two-phase immersion cooling.
• Evaluation: Green HPC systems measure and report the system's energy efficiency and environmental impact using metrics and methods such as power usage effectiveness (PUE), energy-delay product (EDP), and carbon footprint. Green HPC systems also compare and rank the system with other systems, using standards and benchmarks such as the Green500 list.

In summary, green HPC is a holistic approach that considers the entire lifecycle of HPC systems, from design to operation to evaluation, and strives to make them more sustainable and responsible.

Why is Green HPC Important?

Green HPC is important for several reasons, both environmental and economic. By reducing the electricity consumption and carbon emissions of HPC, green HPC can help mitigate the effects of climate change and protect the environment.

Climate change is one of the most urgent and complex challenges of our time, and it poses serious threats to natural ecosystems, human health, and social stability.

According to the Intergovernmental Panel on Climate Change (IPCC), the global average temperature has risen by about 1.1°C since the pre-industrial era. It is likely to reach 1.5°C between 2030 and 2052 if the current trends continue, leading to more extreme weather events, sea level rise, biodiversity loss, food insecurity, and human displacement.

Using green HPC, researchers and academics can contribute to the global efforts to reduce greenhouse gas emissions and limit warming to 1.5°C, as agreed by the Paris Agreement.

Green HPC can also bring economic benefits to the community and society. By lowering the operational costs and increasing the availability and reliability of HPC, green HPC can make HPC more accessible and affordable for research and academic purposes.

HPC is a valuable resource for advancing scientific and scholarly knowledge, and it can also generate social and economic impacts through innovation and collaboration. For example, HPC can help develop new drugs and vaccines, improve weather forecasting and disaster management, optimize energy production and distribution, and enhance education and learning.

However, HPC is also a costly and scarce resource, often limited by the availability of electricity, cooling, and space. According to a report by Hyperion Research, the global HPC market was worth $39.2 billion in 2020, and it is expected to grow to $55.7 billion by 2024.

But the report also estimates that the electricity cost of HPC will increase from $2.7 billion in 2020 to $3.8 billion in 2024, and the carbon emissions of HPC will increase from 3.9 million metric tons in 2020 to 5.4 million metric tons in 2024.

By using green HPC, researchers and academics can reduce these costs and emissions, and make HPC more sustainable and efficient.

Green HPC Technologies

Green HPC technologies are the hardware, software, and system-level solutions that enable green HPC systems to achieve high performance and energy efficiency. There are many green HPC technologies available or under development, and they can be categorized into three main groups:

Hardware: These are the physical components of the HPC system, such as processors, memory, storage, interconnects, and power supplies. Hardware technologies for green HPC aim to reduce the system's power consumption and heat generation while increasing the computing capacity and speed. Some examples of hardware technologies for green HPC are:‍

• Low-power processors: These processors have a lower power consumption and heat dissipation than conventional processors but still provide high performance and scalability. Low-power processors can be based on different architectures, such as ARM, RISC-V, or GPU. For example, NVIDIA DGX SuperPOD is a green HPC system that uses low-power GPU processors to achieve high performance and energy efficiency.‍‍‍
• Liquid cooling: This cooling method uses liquids, such as water or dielectric fluids, to remove heat from the system components. Liquid cooling can be more efficient and effective than air cooling, as liquid has a higher heat capacity and thermal conductivity than air. Liquid cooling can also reduce the noise and dust levels of the system. For example, DUG Technology uses liquid cooling to power its green HPC system, transforming the heat from its servers into useful hot water.‍
• Renewable energy sources: These are energy sources that are derived from natural resources, such as solar, wind, hydro, or geothermal. Renewable energy sources can reduce the carbon emissions and environmental impact of the HPC system, as well as the dependence on fossil fuels and grid electricity. Renewable energy sources can also lower operational costs and increase the reliability of the HPC system. For example, the NYU Greene Supercomputer, one of the world's most energy-efficient supercomputers, uses solar panels and fuel cells to power its green HPC system.

Software: These are the programs and tools that run on the HPC system, such as operating systems, compilers, libraries, schedulers, and applications. Software technologies for green HPC aim to optimize the code and algorithms of the system while adapting to the dynamic and heterogeneous environment. Some examples of software technologies for green HPC are:

• Energy-aware scheduling: This scheduling method considers the energy consumption and performance of the system and allocates the resources accordingly. Energy-aware scheduling can balance the trade-offs between performance and energy efficiency and improve the utilization and availability of the system. EAR is an example of an energy management framework that integrates with SLURM to provide per-job energy consumption accounting and energy-aware scheduling.
• Dynamic voltage and frequency scaling: This technique adjusts the processor's voltage and frequency according to the workload and power budget. Dynamic voltage and frequency scaling can reduce the processor's power consumption and heat generation while maintaining or improving performance and quality. MLPerf is a benchmark suite that measures machine learning applications' performance and energy efficiency and uses dynamic voltage and frequency scaling to optimize the results.
• Edge computing and IoT: These technologies enable the processing and analysis of data at the edge of the network rather than at the central server or cloud. Edge computing and IoT can reduce the data transfer and latency of the system and improve the security and privacy of the data. Edge computing and IoT can also leverage the distributed and heterogeneous resources of the system, such as sensors, devices, and gateways. For example, Atos and HDF Energy are developing a green HPC solution that uses edge computing and IoT to predict the energy consumption needs of the system and optimize it to green hydrogen.

System-level: These solutions integrate and coordinate the hardware and software components of the HPC system, such as cooling, ventilation, monitoring, and control. System-level solutions for green HPC aim to improve the system's overall energy efficiency and environmental impact while ensuring scalability and reliability. Some examples of system-level solutions for green HPC are:

• Two-phase immersion cooling: This cooling method submerges the system components into a dielectric fluid that boils and condenses in a closed loop. Two-phase immersion cooling can provide superior cooling performance and energy efficiency than liquid or air cooling, eliminating the need for fans, pumps, and chillers. Two-phase immersion cooling can also increase the lifespan and density of the system components. Microsoft is testing two-phase immersion cooling for its green HPC system, which uses a fluorocarbon-based fluid with low global warming potential.
• Power usage effectiveness: This metric measures the energy efficiency of the HPC system by dividing the system's total power consumption by the power consumption of the computing equipment. Power usage effectiveness can indicate how much of the power is used for computing and how much is used for other purposes, such as cooling and lighting. The ideal value of power usage effectiveness is 1, which means that all the power is used for computing. The Green500 list ranks the greenest supercomputers in the world based on their power usage effectiveness and performance.
• Carbon footprint: This measures the environmental impact of the HPC system by estimating the amount of greenhouse gas emissions generated by the system. Carbon footprint can reflect the direct and indirect emissions of the system, such as the emissions from electricity generation, transportation, and waste disposal. Carbon footprint can help assess the contribution of the HPC system to climate change and the potential for mitigation and offsetting. For example, the Climate Neutral Data Centre Pact is an initiative that aims to make European data centers climate-neutral by 2030 by reducing their carbon footprint and increasing their use of renewable energy sources.

Green HPC technologies are the hardware, software, and system-level solutions that allow green HPC systems to achieve high performance and energy efficiency. These technologies can have various uses and advantages for different domains of research and academia, such as physics, chemistry, biology, medicine, engineering, and social sciences.

Green HPC in Research and Academia

Green HPC is not only a technical challenge but also an opportunity for research and academia to accelerate their discoveries and impacts. By using green HPC, researchers and academics can leverage the high performance and energy efficiency of HPC systems while reducing their environmental and economic costs.

Green HPC can enable new scientific and scholarly insights, innovation, and collaboration across various domains, such as physics, chemistry, biology, medicine, engineering, and social sciences. Here are some examples of how green HPC is being used in research and academia:

• Physics: Green HPC can help physicists simulate and understand the fundamental laws of nature, such as the origin of the universe, the structure of matter, and the behavior of particles. For example, the Large Hadron Collider (LHC) at CERN uses green HPC to analyze the massive amounts of data generated by its experiments and to discover new particles and phenomena, such as the Higgs boson. The LHC uses liquid cooling and renewable energy sources to power its green HPC system, which is one of the most powerful and energy-efficient supercomputers in the world.
• Chemistry: Green HPC can help chemists design and synthesize new molecules and materials, such as drugs, catalysts, and nanomaterials. For example, the Materials Project at Berkeley Lab uses green HPC to create a database of millions of materials and their properties and to predict the best candidates for various applications, such as batteries, solar cells, and sensors. The Materials Project uses low-power processors and energy-aware scheduling to optimize its green HPC system, one of the most comprehensive and accessible resources for materials science.
• Biology: Green HPC can help biologists model and explore the complexity and diversity of life, such as the evolution of species, the function of genes, and the interaction of ecosystems. For example, the Tree of Life Project at the University of Arizona uses green HPC to construct a phylogenetic tree of all living organisms and to reveal the history and relationships of life on Earth. The Tree of Life Project uses edge computing and IoT to collect and process data from various sources, such as DNA sequences, fossils, and images.
• Medicine: Green HPC can help medical researchers and practitioners diagnose and treat diseases like cancer, diabetes, and COVID-19. For example, the Folding@home project at Stanford University uses green HPC to simulate the folding and misfolding of proteins and to discover new drugs and vaccines. The Folding@home project uses a distributed and volunteer-based green HPC system, which harnesses the idle computing power of millions of devices worldwide.
• Engineering: Green HPC can help engineers design and optimize new systems and products, such as cars, planes, and buildings. For example, the ExaWind project at the National Renewable Energy Laboratory uses green HPC to simulate the aerodynamics and performance of wind turbines and improve the efficiency and reliability of wind energy. The ExaWind project uses two-phase immersion cooling and power usage effectiveness to reduce the energy consumption and carbon footprint of its green HPC system, one of the most advanced and scalable platforms for wind energy research.
• Social Sciences: Green HPC can help social scientists analyze and understand the behavior and interaction of individuals and groups, such as the dynamics of markets, politics, and culture. For example, the World Community Grid at IBM uses green HPC to support various social and humanitarian projects, such as mapping poverty, predicting climate change, and educating children. The World Community Grid uses a cloud-based and donation-driven green HPC system, which utilizes the spare computing capacity of hundreds of thousands of volunteers.

In summary, green HPC is used in research and academia to accelerate discoveries and impacts in various domains. Green HPC can enable new scientific and scholarly insights, innovation, and collaboration across disciplines and sectors.

How to Implement Green HPC

Implementing green HPC in research and academic settings can be challenging but rewarding. It requires careful planning, evaluation, and optimization of the HPC system and the collaboration and coordination of various stakeholders, such as researchers, IT staff, administrators, and vendors.

Here are some practical tips and strategies for implementing green HPC in your own projects and settings:

Assess your needs and goals: Before building or choosing a green HPC system, you should first assess your needs and goals for your research and academic projects. You should consider the following questions:

• What are your projects' main objectives and requirements, such as the type, size, and complexity of the data and computations, the expected performance and quality, and the budget and timeline?
• What are your projects' current and future challenges and opportunities, such as the scalability and reliability of the system, availability and affordability of the resources, and environmental and social impacts?
• What are the best practices and standards for your projects, such as the metrics and methods for measuring and reporting the performance and energy efficiency of the system, the benchmarks and comparisons for evaluating and ranking the system, and the policies and regulations for ensuring the sustainability and responsibility of the system?

Choose the right platform and tools: After assessing your needs and goals, you should choose the right platform and tools for your green HPC system. You should consider the following options:

• Building a new data center: This option gives you the most control and flexibility over the design, operation, and evaluation of your green HPC system, but it also requires the most investment and expertise. You should choose this option if you have a large and long-term demand for green HPC and the resources and skills to build and maintain your own data center.
• Refurbishing an existing data center: This option allows you to improve your existing HPC system's energy efficiency and environmental impact by upgrading or replacing the hardware, software, and system-level components. You should choose this option if you have a moderate and medium-term demand for green HPC and the resources and skills to refurbish and maintain your existing data center.
• Using co-location: This option enables you to rent or share the space and infrastructure of a third-party data center, which provides the hardware, software, and system-level components for your green HPC system. You should choose this option if you have a low and short-term demand for green HPC and do not have the resources and skills to build or refurbish your data center.
• Using the cloud: This option allows you to access the services and resources of a third-party data center, which provides the hardware, software, and system-level components for your green HPC system through the internet. You should choose this option if you have a variable and unpredictable demand for green HPC and do not have the resources and skills to build, refurbish, or co-locate your own data center.

Optimize the code and algorithms: Once you have chosen the right platform and tools for your green HPC system, you should optimize the code and algorithms of your research and academic applications to make them more energy-efficient and environmentally friendly. You should consider the following techniques:

• Parallelization: This technique enables you to divide and distribute the computation of your application among multiple processors or nodes to reduce the execution time and energy consumption. You should use parallelization if your application is suitable for concurrent processing and you have access to a parallel or distributed green HPC system.
• Vectorization: This technique allows you to simultaneously perform the same operation on multiple data elements using a single instruction to reduce the number of instructions and energy consumption. You should use vectorization if your application involves arithmetic or logical operations on arrays or vectors and you have access to a green HPC system that supports vector instructions, such as SIMD or GPU.
• Approximation: This technique involves using a more straightforward or less accurate version of the computation of your application to reduce the complexity and energy consumption. You should use approximation if your application can tolerate some errors or uncertainties in the results and you have access to a green HPC system that supports approximate computing, such as low-precision arithmetic or neural networks.

Monitor and analyze the energy consumption and performance: After optimizing the code and algorithms of your research and academic applications, you should monitor and analyze your green HPC system's energy consumption and performance to evaluate and improve its energy efficiency and environmental impact. You should consider the following methods:

• Profiling: This method involves measuring and recording the energy consumption and performance of your green HPC system at different levels of granularity, such as the system, component, or application level. You should use profiling to identify the sources and patterns of energy consumption and performance and compare and rank the options and alternatives for your green HPC system.
• Modeling: This method involves creating and using a mathematical or statistical representation of your green HPC system's energy consumption and performance based on the data and information collected by profiling. You should use modeling to predict and simulate the energy consumption and performance of your green HPC system under different scenarios and conditions and to optimize and tune the parameters and settings of your green HPC system.
• Reporting: This method involves presenting and communicating the energy consumption and performance of your green HPC system, using metrics and methods such as power usage effectiveness, energy-delay product, and carbon footprint. You should use reporting to demonstrate and justify the energy efficiency and environmental impact of your green HPC system and to share and disseminate the results and best practices.

Share and report the results and best practices: Finally, after monitoring and analyzing your green HPC system's energy consumption and performance, you should share and report the results and best practices to foster a culture of sustainability and responsibility among the HPC community. You should consider the following strategies:

• Collaboration: This strategy involves working and cooperating with other researchers and academics who use or are interested in green HPC to exchange and integrate the data, information, and knowledge of your green HPC system. You should use collaboration to enhance and enrich your research and academic projects and to create new opportunities and synergies for green HPC.
• Publication: This strategy involves publishing and disseminating the results and best practices of your green HPC system in various formats and platforms, such as journals, conferences, blogs, and social media. You should use publication to increase the visibility and impact of your research and academic projects and raise the awareness and interest of the HPC community and society for green HPC.
• Education: This strategy involves teaching and learning about the concepts and techniques of green HPC in various settings and levels, such as courses, workshops, webinars, and online courses. You should use education to develop and improve the skills and competencies of yourself and others for green HPC, and to inspire and motivate the next generation of researchers and academics.

To sum up, applying green HPC in research and academic settings can be a challenging but worthwhile task. It needs thoughtful planning, assessment, and improvement of the HPC system, as well as the cooperation and organization of different stakeholders.

Following these practical tips and strategies, you can successfully implement green HPC in your projects and settings and accelerate your discoveries and impacts.

Green HPC is an emerging field that aims to address the challenges and opportunities of HPC in a more energy-efficient and environmentally friendly way.

Green HPC can reduce its electricity consumption and carbon emissions while maintaining or improving its performance and quality.

It can also lower the operational costs and increase the availability and reliability of HPC, making it more accessible and affordable for research and academic purposes.

Green HPC is not only a technical challenge but also an opportunity for research and academia to accelerate their discoveries and impacts.

By using green HPC, researchers and academics can leverage the high performance and energy efficiency of HPC systems while reducing their environmental and economic costs.

Green HPC can enable new scientific and scholarly insights, as well as innovation and collaboration across various domains, such as physics, chemistry, biology, medicine, engineering, and social sciences.

We hope this article has given you a comprehensive and informative overview of green HPC and its importance in research and academia. If you are interested in learning more about green HPC and how it can help you accelerate your discoveries, contact us at +1 (302) 587-7867, Monday through Friday, 9:00am - 5:00pm EST, or via social media.

Sesterce offers a comprehensive range of services for high-performance computing power using sustainable energies. We support the most cutting-edge industries that use computing power. We are at the center of machine learning and artificial intelligence.

We look forward to hearing from you soon.