AWS re:Invent 2025 - Shell's HPC Evolution: Accelerating Seismic Processing with AWS GPU Innovation

AWS re:Invent 2025 - Shell's HPC Evolution: Accelerating Seismic Processing with AWS GPU Innovation

Overview

• Shell, a global energy company, has been on a multi-year journey to transform its High-Performance Computing (HPC) capabilities by leveraging the AWS cloud.
• This presentation by Michael Gujerol, the Head of Global HPC Engineering and Operations at Shell, and Hussein Sha, the CTO for AWS Energy and Utilities, showcases Shell's HPC evolution and the key challenges, solutions, and benefits they have experienced.

Shell's HPC Challenges and Needs

• Shell's upstream operations, particularly seismic processing, require massive computational power to handle petabyte-scale data and complex geophysical algorithms.
• The need for more compute power has been driven by the increasing complexity of seismic imaging, as Shell explores new energy basins and seeks to improve the accuracy of subsurface models.
• Historically, Shell has faced challenges in quickly scaling and provisioning HPC resources to meet changing business demands, as well as managing the logistics and infrastructure of on-premise HPC systems.

Shell's Journey with AWS

• Shell's journey with AWS HPC began in 2017, but the initial years were marked by slow progress and challenges in replicating their on-premise HPC workflows in the cloud.
• A pivotal moment came in 2021 when Shell set a "10x challenge" - to achieve a 10x improvement in wall-clock time for their seismic processing workloads on AWS.
• This shift in mindset, coupled with advancements in GPU hardware (e.g., Nvidia's P4DE and A100 GPUs) and the availability of affordable high-speed connectivity, enabled Shell to achieve significant performance gains.

Technical Architecture and Workflow

• Shell's hybrid HPC architecture leverages a combination of on-premise resources and AWS services:
  • Petabyte-scale seismic data is first preprocessed on-premise using CPU-intensive tasks.
  • The processed data is then transferred to AWS via high-speed direct connect, where GPU-accelerated full waveform inversion (FWI) algorithms are run on AWS instances (e.g., P5 VMs, Parallel Compute Service).
  • The output data is then transferred back to on-premise for any remaining CPU-based post-processing.
• Shell has also adopted AWS services like FSx for data management and PCS (Parallel Compute Service) for seamless Slurm-based job orchestration.

Business Impact and Benefits

• By leveraging the scalability and flexibility of AWS, Shell has been able to achieve 3-5x acceleration in their seismic processing workloads, compared to their on-premise HPC systems.
• This has resulted in a cumulative 2.5 years of wall-clock time savings over the past 1.5 years, enabling Shell to make faster decisions and accelerate their upstream operations.
• The ability to rapidly provision and scale HPC resources on-demand has also transformed Shell's internal workflows and mindsets, allowing their researchers and engineers to experiment and innovate more freely.

The Road Ahead

• Shell's focus for the future includes:
  • Further integration of HPC workflows with other data and AI/ML services in the AWS ecosystem, to create a more seamless and efficient end-to-end solution.
  • Achieving true multi-region HPC capabilities to enhance resilience and flexibility.
  • Continued optimization and exploration of the latest GPU hardware and cloud services to push the boundaries of price-performance.
  • Fostering deeper collaboration with AWS and the broader industry to develop innovative, cloud-native HPC solutions and best practices.

Key Takeaways

• Shell's journey demonstrates the transformative potential of cloud-based HPC for energy companies, enabling them to overcome the scalability, flexibility, and cost challenges of traditional on-premise HPC.
• By embracing a "cloud-first" mindset and focusing on achieving significant performance gains, Shell was able to unlock new ways of working and accelerate their critical seismic processing workflows.
• The partnership between Shell and AWS highlights the importance of co-innovation and industry collaboration in driving the adoption of cloud-based HPC solutions tailored to the unique needs of the energy sector.