REINVENTING ENERGY

High Performance
Computing (HPC)
Table of
AWS for Energy and Utilities: Reinventing the energy industry 3

Energy Systems of the Future: The tools for navigating the road ahead 4

Contents The Growing Need for HPC and Current Challenges:

Optimizing data value for faster, better decision making
5

The AWS Difference: Unlimited compute power when you need it 6

AWS HPC in Action:

Shell: Thinking hybrid helps optimize performance and minimize costs 8

Baker Hughes: Reducing time to results, carbon footprint and cost 10

Woodside Energy: Transforming the speed and accuracy of seismic surveys 12

BlocPower: Optimising the speed, cost and flexibility of energy data processing 14

Next Steps: Let’s talk HPC, now

16
 HPC
ABOUT AWS ENERGY AND UTILITIES

AWS for Energy and Utilities

These energy systems of the future, built with
Reinventing the energy industry. AWS’s deepest and broadest cloud services will
result in energy systems that are more:
Amazon Web Services (AWS) is the world’s most comprehensive and broadly adopted cloud,
offering over 200 fully featured services from data centers globally. Millions of customers – including Resilient
the fastest-growing startups, largest enterprises, and leading government agencies – are using AWS Handles critical events, including
increasingly volatile climate and weather,
to lower costs, become more agile, and innovate faster.
new consumption patterns and consumer
With today’s energy industry rapidly changing, the need This ebook, focused on High Performance Computing (HPC), expectations, regulatory revisions, supply
for energy companies to innovate is critical. As customer is part of an ongoing series that will dive deep into the critical chain disruptions, and economic volatility.
behaviors evolve, energy demand increases, and the need cloud technologies that are helping energy companies across
to decarbonize accelerates. Energy organizations will need the value chain reinvent the industry as they build the energy
Responsive
to increase innovation through the latest cloud technologies systems of the future.
Senses and adapts in real-time to deal
such as AI, ML, edge computing, and many others.
with unexpected changes in the operating
environment or respond quickly to emergencies.

Resourceful
Maximizes existing assets, people and information,
which leads to improved operational excellence
and financial performance.

Reliable
Scales across a large footprint sustainably,
while ultimately delivering safe, reliable
energy to everyone who depends on it.

Renewable
Meets today’s demand while making the shift
to renewable energy sources and contributing
to a balanced global energy mix.

3
 HPC
INTRODUCTION

The Energy Systems of the Future The energy systems of the future allow you to put your data
to work and provide you the tools to uncover insights that will

The tools for navigating the road ahead. help optimize energy operations and accelerate innovation.
These data-driven solutions are based on:

Advanced High-performance
Working together with an expansive network of energy partners, customers and startups, AWS analytics computing
is building the energy systems of the future and helping leading organizations across the value
chain increase energy production, at lower cost, with lower risks, and lower carbon emissions, Artificial intelligence/ Next-generation
Machine learning connectivity solutions
safely and securely, for generations to come.
Internet of Things/
Edge computing

AWS is reinventing the energy industry through innovative

cloud technologies and comprehensive digital solutions for:

Oil and gas The modernized, Clean and renewable

exploration scalable energy technologies*
and production electrical grid

*Such as wind, solar, battery storage, clean hydrogen

generation and carbon catpure and sequestration.

In the chapters ahead, we look at how

high-performance computing (HPC) on
AWS is transforming subsurface exploration
and other parts of the energy industry,
through faster processing, greater
availability, and unlimited capacity.

4
 HPC
THE NEED FOR ACTION

The Growing Need for HPC

and Current Challenges
Optimizing data value for faster, better decision making.

The biggest oil and gas operators have been building and maintaining growing HPC clusters
for seismic processing. Seismic HPC is used to crunch the data to reduce uncertainty and increase
the quality of subsurface decisions.

The seismic HPC estate is traditionally siloed from the rest The other main HPC workflow is related to reservoir
of the IT infrastructure. Projects from asset teams are queued simulation. This is often relegated to individual workstations
to be processed through the HPC cluster. As the seismic HPC or departmental HPC servers or clusters, that are not sized to
is a finite resource, it is not unsual for compute jobs to be deliver the scale the business requires. The Future of HPC
‘bumped’ to make space for jobs with higher priority.
One of the largest new and up and coming HPC workloads is The limitations of on premise HPC in terms of availability,
Wait time can vary depending on the number of workloads related to Artificial Intelligence (AI)/Machine Learning (ML) and capacity, flexibility, and investment required have restricted
and their level of priority. But it can run into months, frustrating Large Language Modelling (LLM). Almost no energy company its application across the energy industry.
staff and holding organizations back. More HPC capacity can is fully equipped with the infrastructure required for these kind
have strong competitive advantage, especially with very large of workloads, and most AI/ML is native to the cloud, both in At AWS, we believe in applying best practice wherever
offshore projects. terms of hardware and software tools, and ecosystem. possible and therefore extend the definition of HPC into
technical and scientific computing. In addition to traditional
Medium/small operators outsource seismic processing, Leveraging AWS to complement existing HPC infrastructure HPC, we include AI/ML, 3D visualisation and more, where
but they would benefit from some in-house HPC for the provides more flexibility and adaptability to business needs. AWS can deliver major benefits.
imaging and interpretations portions of the workflow where This is just one of the many HPC challenges that AWS for
the geoscientists are testing new hypotheses and this happens Energy solves for the entire energy industry, outlined in This is our vision for HPC, enabled through our ability
interactively, or an outsourced model is slow, due to the the following pages. to optimize availability and capacity.
procurement process and the time to results.

5
 HPC
THE HPC ADVANTAGE

The AWS Difference

Unlimited compute power when you need it.

Here’s how AWS for Energy is shifting the dial in HPC, overcoming the barriers presented by on-
premise systems, accelerating workload processing and opening the technology to more business
applications:

Adapting to business needs Providing the best HPC for all workloads
Extending on-premise HPC with AWS HPC, Typically, on-premise HPC infrastructure is
can reduce or completely remove the waiting specifically designed for the most business-critical
process. All HPC workloads can be handled workloads (often seismic HPC). Frequently, other
quickly and simultaneously, reducing time for workloads are not designed to take full advantage
even the most compute-demanding jobs to days of the same hardware architectures, driving
or weeks rather than months. There is no longer companies to compromise. These limitations don’t
a wait cycle dictated by limited HPC resources. exist in the AWS Cloud. Each workload benefits
from the most appropriate infrastructure to meet
Supporting innovation and research its needs when and where it is required.

With computing capacity available whenever

it’s required, not only is queuing a thing of
Increasing software agility
the past with AWS HPC, but also there’s no Traditionally, businesses buy highly specialized
need for research and other less time-critical on-premise HPC hardware and optimize their
workloads to cede priority to those considered workflows and algorithms around these restricted
more important. HPC is always available, specifications. One of the main advantages of AWS
whatever the workload, accelerating research is that it is natively heterogeneous, offering all
and innovation. Practically all the new AI/ML possible compute architectures, including x86, GPUs,
workloads are in ‘research and innovation’ ARM and even new specialized architectures for AI/
mode, and it is no coincidence that most of ML learning and inference. This makes it possible
these workloads are cloud-native. (and highly desirable) to diversify how algorithms
or calculations are developed and optimized. The
result is modern compute software that can choose
between costs, speed and availability of resources.

6
 HPC
THE HPC ADVANTAGE

The AWS Difference (continued)

Delivering scalability and efficiency Enabling global connectivity

The very large capital investment required for HPC is applied to large data sets, which are
on-premise HPC forces organizations to keep acquired from one or more locations and systems
the HPC infrastructure fully utilized all the time. beyond where the HPC is taking place. The AWS
This drives compromises on the spending allowed. Cloud is a natural collaboration hub, making it easy
With AWS, capacity is available as and when for data to be collected, shared and transferred
needed, and customers are only charged for across the globe between teams inside the same
what they use, optimizing cost and efficiency, and company and between third parties collaborating
reducing constraints for IT and the wider business. on the same project.

Accelerating access to new chipsets This capability will be soon augmented by AWS
Project Kuiper, connecting the AWS on-planet
Microprocessor vendors are on a 12 to 18-month
network with a very large and highly efficient
innovation cycle, while the procurements
network of off-planet satellites. The integration
process to buy and implement a new large
and acceleration of data sharing from collection
on-premise HPC data center is 9 to 12 months.
and processing to visualization and decision
Plus, the energy industry is competing for new
making that this delivers saves months, speeds
hardware that is now in high demand in other
up the decision making process and avoids costly
less price-sensitive industries, which drives up
operational mistakes.
costs and long wait times.

AWS is one of the largest buyers for all

microprocessors, and new systems are constantly
becoming operationally available on AWS months Of course, each energy company has
before the same solution can be implemented its own specific HPC challenges.
on-premise. The scale and flexibility of HPC
resources at AWS enables faster access to the Over the following pages you’ll find out
necessary HPC hardware when required and how AWS has been helping four businesses
enables the energy industry to stay in line with overcome them and drive success.
technology innovation and the price/performance
advantages introduced by new chipsets.

7
 HPC

AWS HPC IN ACTION

Shell
Thinking hybrid helps optimize performance and minimize costs.

The international energy giant has expertise in the exploration, production,

refining and marketing of oil and natural gas, and the manufacturing and
marketing of chemicals, with operations in more than 70 countries.

The challenge to consume over 80% of all compute cycles. Successfully

accelerating the process will have a major impact on Shell’s
The geophysics team at Shell started their cloud journey with
overall goal.
AWS five years ago with the idea of moving all of their HPC to
AWS with a pure on-demand model in mind. While this was
A hybrid model drives the FWI solution. Shell’s on-premise
technically feasible, it faced several challenges.
infrastructure continues to handle the rest of the workflow,
while FWI is performed on AWS GPUs. One of the enabling
After working closely with AWS on several proof of concepts,
technologies has been AWS Direct Connect. Shell has two
the team decided to think much bigger to see just how far they
10Gbit lines connecting their on-premise HPC to AWS,
could push forward their end-to-end workflows and seismic
enabling them to transfer 30TB overnight. The results are
processes in the cloud. Although the results were encouraging,
then transferred back to Shell on-premise. As the project
it proved difficult to optimize the bespoke complete workflows.
progresses, the rest of the Seismic Interpretation workload
With seismic data handling increasingly needing more
will move to AWS and that’s also where the data will remain.
computing power, the team knew the future lay in the cloud,
and began identifying a better approach. They set an end goal
The compute density of the AWS Nvidia GPUs enables Shell
of accelerating time to decisions by 10 times over several years
to achieve very large compute capacity with a fraction of the
by leveraging multiple technical innovations.
servers compared to CPU-based systems. Shell has dedicated
compute capacity using a few P4de EC2 instances, plus it can
The solution add on-demand capacity on a per-project basis when needed.
In 2022, the project pivoted to focus on Full-Waveform While today an FWI job can consume 150 P4de systems (1,200
Inversion (FWI) on GPUs as the target algorithm. FWI is the most Nvidia A100 GPUs) for a couple of weeks, it is projected to grow
strategic long-term solution in seismic imaging and is forecast up to five times over the next two to three years.

8
 HPC
AWS HPC IN ACTION

Shell (continued)

The outcome
This hybrid solution has given Shell’s geophysics team the flexibility to burst into bigger capacity in the cloud as and when different
projects demand it, while retaining access to the bespoke nuances of an on-premise system matured over 40 years. This optimizes
performance, while minimizing costs as Shell only pays for the HPC capacity it uses.
“
We still have a reasonably
The focus on FWI is also driving joint innovation across the Shell code and architecture, AWS compute and storage capabilities, sized on-premise data
center for CPU and
and even driving Shell and AWS to create an industry platform that will enable commercial and proprietary software to coexist,
delivering a best-of-breed ecosystem.

Key fact storage. It’s what’s most

cost effective right now.
As we get better at seismic
2 x 10Gbit processing in the cloud,
lines connect Shell with AWS Cloud so…
and it becomes more
commercially attractive,
30TB then it will probably all
of data can be transferred over night go to the cloud.”
Tim Roden
General Manager, Geophysics, Western Hemisphere, Shell

9
 HPC

AWS HPC IN ACTION

Baker Hughes
Reducing time to results, carbon footprint and cost.

One of the world’s largest oil field services companies, Baker Hughes delivers products,
technology, and systems to the oil and natural gas sector across 120 countries.
It also provides downstream chemicals, and process and pipeline services.

The challenge The solution

Companies across the energy sector rely on Baker Hughes’ Working with AWS Professional Services to quickly develop a
NovaLT gas turbines, which have set new standards in proof of concept that would optimize simulation runtime and
greenhouse gas emissions, efficiency and reliability. performance, Baker Hughes migrated its computational fluid
Running effective simulations is key to improving dynamics (CFD) applications to the AWS Cloud.
and evolving the design of these essential products.
Using Amazon Elastic Compute Cloud (EC2), secure and
However, engineers at Baker Hughes’ Turbomachinery scalable capacity is now available for virtually any workload.
and Process Solutions (TPS) Research Centre in Italy were HPC is accelerated only when necessary using Elastic Fabric
experiencing resource capacity bottlenecks. This was Adapter (EFA), a network interface that enables Amazon
causing long wait and run times, restricting the number EC2 to run applications requiring high levels of internode
of simulations and increasing the need to run costly and communications at scale.
demanding physical tests.
EFA delivers a dedicated throughput of 100 gigabits per second
The company calculated that fixing the delays and better per HPC job compared to the traditional network interface,
managing peak periods required expanding computing which offers 300 gigabits per second throughput shared across
capacity to 400 teraflops. But it wasn’t financially sustainable multiple HPC jobs. As a result, HPC jobs using EFA have low
for the business to pay for peak capacity all year round. A latency compared to the traditional network interface at a
more flexible, scalable solution had to be found to optimize fraction of a cost.
performance and minimize costs.

10
 HPC
AWS HPC IN ACTION

Baker Hughes (continued)

“
The outcome
More than 150 Baker Hughes TPS engineers in Italy, India, take performance to new heights, the most complex simulations
and the US are now able to run as many simulations as needed with specific memory requirements that could not be handled
prior to physical tests, leading to better accuracy with fewer on premise can now be run on AWS.
test iterations. Plus, multiple users can be onboarded every
month without impacting HPC job performance. Financial savings have also been significant, with HPC spend 40% We were initially
planning to migrate
lower thanks to a range of cost-optimization solutions, including
The most resource-intensive jobs can now be carried out with AWS MAP, Savings Plans, and EDP. Plus, the company has reduced
98% less wait time and 26% faster on AWS compared with the the carbon footprint of its HPC workloads by 99% compared
on-premises HPC solution, accelerating the simulation process.
Engineers can also run design simulations in parallel on AWS
with on premise due to the AWS customer carbon footprint tool,
which uses simple-to-understand data visualizations to help
the equivalent compute
compared with running them sequentially on premise. Helping customers review, evaluate, and forecast emissions. capacity of 100 teraflops
Key facts to AWS, but by giving
engineers the possibility
to scale, the consumption
40% 90% 26% spiked by four times
within three months
HPC cost drop in simulation faster simulation
of go-live.”
reduction wait time run time
Yogesh Kulkarni
Senior Director, CTO India, Baker Hughes

11
 HPC

AWS HPC IN ACTION

Woodside Energy
Transforming the speed and accuracy of seismic surveys.

Australia’s largest independent oil, gas and new energy company, Woodside Energy
recently merged with BHP Petroleum. Its goal is to provide low-cost, low-carbon energy.

The challenge The solution

Full waveform inversion (FWI) is a high-resolution seismic AWS worked from day one with S-Cube, and Woodside became
imaging technique that uses the entire content of seismic traces the first energy company to carry out FWI processing in the
for extracting physical parameters of the medium sampled by cloud. Another industry milestone was achieved by deploying
seismic waves. It can be used to reconstruct high-resolution 1 million vCPUs in parallel in a single FWI job using an S-Cube
subsurface models, increasing the accuracy of seismic surveys. algorithm running on AWS. S-Cube also collaborated with
AWS to take advantage of AWS’s Graviton2, the first ARM
Woodside Energy, one of Australia’s leading energy producers, microprocessor available in the cloud, with a very strong
had spent several years working on an advanced FWI technique price-performance advantage and a very low energy footprint.
to optimize and streamline the survey process, in close
collaboration with S-Cube, a spin-off from London’s Imperial Woodside has also implemented AWS Snowball to transport
College (one of the inventors of FWI). data from a survey while it was still being acquired. This
enabled the processing of the data to start while the seismic
evaluation was taking place, significantly accelerating the
development of an advanced FWI technique.

12
 HPC
AWS HPC IN ACTION

Woodside Energy (continued)

The outcome
Woodside was able to deliver a more effective and powerful
version of FWI within just two weeks of the completion of
the survey, significantly increasing the speed and accuracy
Key facts
“
Traditionally we’d
done processing, then
of the process.

Furthermore, this was achieved without the data leaving the

cloud and for a fraction of the cost of carrying out the work
on premise. In fact, the energy company is confident that it
model building, then
will reach a point when costs will be as low as 1% of the original
on-premise process. First time FWI imaging, then inversion.
Together, Woodside, S-Cube and AWS have pushed the
processed in the cloud in the energy sector The opportunity now
boundaries in FWI for an industry which now more than ever
demands accuracy and automation-leveraging productivity is to do all that in
one step and going to
gains and a deeper understanding of the subsurface.

Woodside is now leveraging this solution in more projects

together with the new algorithms from S-Cube. FWI as a final product
– and that’s within
sight because of the
1 million opportunity of cloud
vCPUs deployed in parallel in a single
compute instance – another industry first
compute with AWS.”
Source: https://www.s-cube.com/news/posts/2020/december/
deployment-by-woodside-energy-using-s-cube-algorithm-running-on-aws/
Tony Almond
VP Technology and Innovation, Woodside Energy

13
 HPC

AWS HPC IN ACTION

BlocPower
Optimising the speed, cost and flexibility of energy data processing.

On a mission to decarbonize buildings, BlocPower has successfully implemented electrification,

solar, and other energy efficiency measures in more than 1,200 buildings across the US to date.

The challenge The solution

There are around 130 million buildings across the US, BlocPower discovered AWS Batch, which provides fully managed
and they account for around 30% of the country’s carbon batch data processing at virtually any scale. Using Amazon Elastic
emissions. Climate technology leader BlocPower wanted to Container Service, the company containerized its workloads,
build a powerful, cost-effective data processing pipeline to which were then orchestrated by AWS Batch. This accelerated
process the energy profiles for each of these buildings to data processing speeds by 16,000 times and processed over 30TB
better understand how to optimize their energy efficiency. of data, completing the job in an hour, rather than the thousands
of hours it would have taken on premises.
The company sources its energy profiles using EnergyPlus,
the US Department of Energy’s open-source whole-building To protect data and optimize scalability, BlocPower’s HPC
modelling engine. BlocPower needed to adopt a set of HPC compute environment uses Amazon Elastic Compute Cloud
solutions that would be compatible with its C++ software (Amazon EC2), which provides secure and resizable capacity
development kit. It also faced the problem that data from for virtually any workload. Meanwhile, to keep costs to a
buildings across the US are not compiled in a clean way minimum, diversified Amazon EC2 Spot Instances was adopted,
making it impossible to apply the necessary analytics tools which delivers better price performance for compute-intensive
without first unifying the data. workloads. Data is then visualized using Amazon QuickSight,
a cloud-native, serverless business intelligence service.
Because the company has been cloud native to AWS since
2016, it turned to us to find scalable compute and data
processing solutions that would work alongside the C++
software development kit.

14
 HPC
AWS HPC IN ACTION

BlocPower (continued)

“
The outcome
Using AWS solutions, BlocPower has streamlined
its data management, improved the performance

Our data processing

of its query processes, and can run advanced
analytics that help it gain insights into improving
the energy efficiency of buildings. The collaboration
has also enabled the company to expand its
services across the US.
would have taken
With access to the latest AWS cloud technology,
thousands of hours on
BlocPower will continue launching services that
help it work towards its goal of decarbonizing
premises. Using AWS
buildings and developing an efficient energy
infrastructure in underserved communities. Batch, we can process
Key facts that data in under an
hour, and can scale to
process our data for
a few hundred dollars
16,000 Millions 30TB every month, rather
times increase in data
processing speeds
of dollars saved in
compute costs
of data processed
than millions.
Ankur Garn
Director of Data Architecture and Analytics, BlocPower

15
 HPC

NEXT STEPS

Let’s Talk HPC, Now

The previous four examples only scratch the surface of the value
AWS is bringing to the energy industry by delivering easily accessible,
scalable and unlimited HPC.

The ability AWS HPC gives energy companies across the sector to rapidly process and analyze
huge amounts of data, incorporating the influence of previously overwhelming numbers of
variables, opens the door to advanced simulations. This plays a key role in delivering a consistent,
efficient energy supply safely and securely with minimum environmental impact.

Oil and Gas

Highly accurate seismic and reservoir simulation can pinpoint new resource locations and optimize extraction
efficiency while minimizing energy usage and cost. Optimal processing and transport strategies can also be
visualized ensuring resources are maximized.

Grid Modernization
Future load, demand and extreme weather events such as storms and wild fires can be simulated to help design
and build more robust and resilient grids ensuring utilities are better prepared for the changing enviorment.

Renewables
HPC-driven simulations can inform the design of wind and solar systems that optimize energy production
and delivery. When adding other data variables such as historical weather patterns, simulations provide
optimal layouts for new renewable sites.

To learn how you can get started with HPC today for your
business, contact an AWS Energy expert.
Contact Us

16
 HPC

A majority of the stories and topics shared within

this ebook and throughout the series originated
from our annual AWS Energy Symposium.
This 2-day event takes place in Houston, Texas and gathers AWS experts and leaders from
energy companies across the value chain to discuss how cloud technologies are helping
the industry:

Increase energy
production Reduce emissions

Decrease Maximize safety

operational cost & security

Improve customer
experience

Find out more and hear the energy companies above and others tell their own stories on the
AWS Energy Symposium website, where you’ll discover more key insight to enable your business
to embrace the Energy Systems of the Future and drive success.

Energy Symposium Website

17
www.aws.com