��ҹѰ��

Accelerating Seismic Insight: How HPG from UNICAMP and Azimuta Used ��ҹѰ�� to Scale HPC Workloads in AWS

Delivering faster seismic insights through hybrid cloud bursting, helping UNICAMP accelerate workflows and maximize HPC performance in AWS.

At the — part of UNICAMP, the “Universidade Estadual de Campinas” in Brazil — researchers push the limits of computation to uncover critical seismic insights that drive exploration and innovation in the oil and gas industry. Led by Jorge Henrique Faccipoeri Jr, the lab develops advanced mathematical models and GPU-optimized software to process terabytes of seismic data, revealing subtle subsurface features such as fractures, caves, and traps where oil and gas may be found.

Faccipoeri partnered with Caian Benedicto, a software developer from HPG, and several others to found AZIMUTA GeoSolutions (AZIMUTA), an independent geophysical services company specializing in high-performance seismic processing and imaging solutions for the oil and gas industry. Together, HPG and AZIMUTA focus on advancing high-performance computing (HPC) techniques to accelerate and enhance seismic interpretation.

As workloads and data volumes expanded, the lab’s on-premises infrastructure could no longer meet peak demand. To accelerate research while maintaining flexibility, HPG partnered with ��ҹѰ�� and AWS to explore a cloud bursting solution that seamlessly connected on-prem and cloud environments.

The Challenge

Seismic processing demands both high throughput and parallel performance. Each dataset can span multiple terabytes, with distributed GPU-based software reading data from many compute nodes simultaneously. HPG’s on-premises storage infrastructure worked well for daily workloads, but as multiple departments shared resources, compute and storage availability became constrained.

To keep projects on schedule, the team needed the ability to burst into the cloud on demand, leveraging AWS compute resources without overhauling its existing workflows or duplicating massive datasets. The solution also needed to preserve full visibility between on-premises and cloud storage while providing the same performance profile researchers expected in the lab.

The Solution

Working with ��ҹѰ��, HPG deployed a hybrid configuration that uses ��ҹѰ�� Core on-premises and Cloud Native ��ҹѰ�� (CNQ) on AWS, connected via ��ҹѰ�� Cloud Data Fabric (CDF).

CDF serves as a data bridge between environments, allowing researchers to access and process the same files from either location without migrating or replicating full datasets. Data portals were configured with the on-premises cluster as the hub and the AWS CNQ cluster as the spoke, creating an intelligent caching layer in the cloud.

When researchers launched seismic processing jobs in AWS, ��ҹѰ��’s NeuralCache automatically prefetched only the necessary data blocks. This adaptive caching, powered by machine learning, enabled rapid re-runs of experiments with significantly faster access times.

The Deployment Journey

The proof-of-concept was designed to validate performance, scalability, and ease of integration. Within days, ��ҹѰ�� Core was operational in HPG’s data center, CNQ was deployed in minutes on AWS, and Cloud Data Fabric seamlessly connected the two environments. Once configured, ��ҹѰ��’s data fabric automatically synchronizes data between on-premises and cloud environments, updating only changed data blocks and keeping both environments perfectly aligned.

This intelligent, strongly consistent cache removed the need to manually maintain file copies or reconcile versions, ensuring that every data access in AWS referenced a single source of truth. Researchers could confidently burst compute workloads to the cloud, knowing their applications were always using the most current data without altering existing workflows.

These results demonstrated how ��ҹѰ��’s intelligent caching and consistent file system experience enable HPC workflows to run seamlessly across environments, with no application changes required.

Business Impact

• Faster Time to Insight: Cloud bursting reduced data access latency by up to 3x, accelerating iterative seismic experiments and model tuning.

• Elastic Scalability: Researchers can now scale GPU-accelerated compute resources in AWS on demand, without being limited by local hardware availability.

• Seamless Hybrid Data Access: Cloud Data Fabric delivers a unified data layer across on-prem and cloud, eliminating the need for manual migration or duplication.

• Simplified Infrastructure: Setup was straightforward, and the hybrid configuration required no re-architecture of applications or workflows.

Looking Ahead

Together, ��ҹѰ��, HPG, and AZIMUTA are paving the way for the next generation of hybrid HPC workflows by combining the elasticity of cloud computing with the performance of on-premises systems. With intelligent data caching and seamless cloud bursting, ��ҹѰ�� enables researchers to focus on advancing seismic science rather than managing infrastructure.