What is a Cloud-Based Supercomputer?
A cloud-based supercomputer provides high-performance computing using cloud resources so multiple users can share access while maintaining privacy and security. The architecture of cloud-native supercomputers manages computing and communications in parallel for enhanced efficiency. This efficiency is achieved by using three complimentary types of processors.
- CPU - Central processing units (CPUs) provide the fast serial processing that algorithms require. Their efficiency can be challenged by the complexity of supercomputing tasks. CPU functionality is augmented with two additional kinds of processors.
- DPU - Data processing units manage provisioning, virtualization, and hardware utilization. DPUs enable supercomputing nodes to provide bare-metal performance and multi-tenancy. A DPU lets the CPU offload routine tasks so it can concentrate on processing for improved performance.
- GPU - Graphics processing units act as co-processors and improve application speed by running multiple, parallel searches.
Combining the processing power of supercomputers with the scalability and connectivity of the cloud opens the door to innovative business and scientific solutions. Traditional supercomputers were out of reach for most companies. The cloud has brought supercomputing to the mainstream and promises to change the way businesses perform research and development to deliver complex products.
Examples of Cloud-Based Supercomputer Offerings
Many cloud vendors have added supercomputing offerings to their portfolio of services. They bring the power of supercomputing to organizations of any size that could never afford to support an in-house environment.
- Google Cloud’s high-performance computing (HPC) service lets companies get started quickly with easy access to a cost-optimized infrastructure designed to foster innovation. Users only pay for the resources they need and can take advantage of sustained-use discounts. Low-latency and high-throughput object and file storage provide exceptional performance.
- NVIDIA's Cloud-Native Supercomputing offering provides a secure HPC platform with enhanced data protection and multi-tenant isolation. The service delivers optimal bare-metal performance with increased efficiency and reduced system noise.
- Amazon’s HPC offering provides users with a suite of products and services designed to deliver faster insights and lets businesses move quickly from development to the market. It’s powered by high-performance file systems and networking solutions.
- Microsoft Azure’s HPC solution offers customers a complete set of computing, networking, and storage resources integrated with workload orchestration services. The service includes machine learning tools for smarter simulations and enhanced decision-making. Azure’s supercomputer has been recognized as one of the fastest in the world.
Supercomputers are no longer a luxury to be enjoyed by organizations with deep pockets. The examples above are just a few of the options available to customers who desire the power of a supercomputer. Cloud-based supercomputing is another example of how effective use of the cloud can change everything.