Executive Summary

To enhance its research capabilities and support data-intensive academic workloads, the American University of Sharjah (AUS) deployed a High-Performance Computing (HPC) environment on Amazon Web Services (AWS). In partnership with Zero&One, AUS implemented a cloud-native solution designed to accelerate scientific simulations, artificial intelligence (AI) training, and advanced data modeling.

Leveraging AWS ParallelCluster, Amazon FSx for Lustre, and Amazon SageMaker, the new environment delivers scalable compute and GPU resources that adapt dynamically to research demands. Zero&One architected the solution following AWS Well-Architected best practices, emphasizing security, automation, and cost optimization.

Through continuous monitoring, performance tuning, and managed operations, the project empowers AUS researchers to focus on innovation while benefiting from a secure, efficient, and fully managed HPC infrastructure on AWS.

Customer Challenge

The American University of Sharjah (AUS) faced increasing computational demands from its expanding research programs in computational mathematics, scientific computing, and artificial intelligence. The university's existing on-premises infrastructure lacked the flexibility and scale needed to support complex simulations and GPU-intensive workloads.

Researchers required a modern environment capable of running large-scale numerical simulations, finite element and finite difference models, and deep learning training using frameworks like MATLAB, Julia, and PyTorch. The absence of on-demand scalability limited their ability to process massive datasets efficiently, often leading to delays and resource bottlenecks.

Additionally, AUS needed to ensure that its research workloads met strict security, compliance, and data governance requirements, while maintaining visibility and control over cloud resource utilization and costs. The university sought a cloud-based, secure, and cost-efficient HPC environment that could accelerate research innovation without adding operational overhead.

Why Zero&One and AWS

AUS partnered with Zero&One, an AWS Premier Consulting Partner, for its strong track record in delivering secure, high-performance cloud environments for education and research. Leveraging the scalability and flexibility of Amazon Web Services (AWS), Zero&One architected a tailored HPC solution that combines automation, cost efficiency, and advanced security to meet AUS's evolving research and compliance needs.

The Solution

To address the computational and scalability challenges faced by AUS, Zero&One designed and deployed a technically advanced High-Performance Computing (HPC) architecture on Amazon Web Services (AWS), fully aligned with the AWS Well-Architected Framework.

The solution centers on AWS ParallelCluster, which automates provisioning, configuration, and scaling of compute environments for both CPU- and GPU-intensive workloads. A Head Node hosts the SLURM scheduler and serves as the administrative gateway via AWS Systems Manager (SSM) or NICE DCV, managing user sessions and job submissions securely.

Dynamic Compute Nodes—optimized for memory, CPU, or GPU performance—are automatically scaled based on research workloads, ensuring maximum performance and cost efficiency. Data management is handled by Amazon FSx for Lustre, which provides low-latency, high-throughput storage, tightly integrated with Amazon S3 for efficient data movement using lazy-loading mechanisms.

For visibility and control, Prometheus and Grafana deliver real-time performance metrics, while Amazon CloudWatch, AWS CloudTrail, and AWS Config centralize logging, compliance tracking, and auditing. Security is enforced through AWS Identity and Access Management (IAM), AWS Security Hub, and Web Application Firewall (WAF), ensuring granular access control and proactive threat detection.

Networking is built within a dedicated VPC, segmenting the environment into public and private subnets. A NAT Gateway allows secure outbound connectivity for private compute resources without exposing them to the internet.

This architecture empowers AUS researchers to execute complex Finite Element Method (FEM) and Finite Difference Method (FDM) simulations, as well as deep-learning workloads using PyTorch and SageMaker, with the flexibility to scale compute power on-demand while maintaining high levels of security and governance.

Results & Benefits

The deployment of the AWS High-Performance Computing (HPC) environment at the American University of Sharjah (AUS) delivered major improvements in research speed, scalability, and efficiency. By adopting AWS ParallelCluster with auto-scaling compute nodes, AUS achieved up to 60% faster processing for simulations and a 3x increase in compute capacity during peak workloads.

Through continuous monitoring and cost governance using AWS Budgets and nOps, the university optimized spending by 25-30% while maintaining 99.9% uptime under Zero&One's 24/7 Managed Services. Researchers now access resources on demand, enabling faster experimentation and reduced queue times for AI and scientific workloads.

The integration of Amazon SageMaker and GPU-powered instances expanded AUS's capabilities in data science and machine learning, while the application of AWS Well-Architected best practices enhanced security and compliance. Overall, AUS can now focus on advancing innovation rather than managing infrastructure, accelerating research outcomes with greater agility and cost control.