Business Software for AWS Nitro System

Amazon Web Services (AWS) is the world’s most comprehensive cloud platform, trusted by millions of customers across industries. From startups to global enterprises and government agencies, AWS provides on-demand solutions for compute, storage, networking, AI, analytics, and more. The platform empowers organizations to innovate faster, reduce costs, and scale globally with unmatched flexibility and reliability. With services like Amazon EC2 for compute, Amazon S3 for storage, SageMaker for AI/ML, and CloudFront for content delivery, AWS covers nearly every business and technical need. Its global infrastructure spans 120 availability zones across 38 regions, ensuring resilience, compliance, and security. Backed by the largest community of customers, partners, and developers, AWS continues to lead the cloud industry in innovation and operational expertise.

Amazon Elastic Compute Cloud (Amazon EC2) is a web service that provides secure, resizable compute capacity in the cloud. It is designed to make web-scale cloud computing easier for developers. Amazon EC2’s simple web service interface allows you to obtain and configure capacity with minimal friction. It provides you with complete control of your computing resources and lets you run on Amazon’s proven computing environment. Amazon EC2 delivers the broadest choice of compute, networking (up to 400 Gbps), and storage services purpose-built to optimize price performance for ML projects. Build, test, and sign on-demand macOS workloads. Access environments in minutes, dynamically scale capacity as needed, and benefit from AWS’s pay-as-you-go pricing. Access the on-demand infrastructure and capacity you need to run HPC applications faster and cost-effectively. Amazon EC2 delivers secure, reliable, high-performance, and cost-effective compute infrastructure to meet demanding business needs.

Amazon Elastic Compute Cloud (EC2) Trn1 instances, powered by AWS Trainium chips, are purpose-built for high-performance deep learning training of generative AI models, including large language models and latent diffusion models. Trn1 instances offer up to 50% cost-to-train savings over other comparable Amazon EC2 instances. You can use Trn1 instances to train 100B+ parameter DL and generative AI models across a broad set of applications, such as text summarization, code generation, question answering, image and video generation, recommendation, and fraud detection. The AWS Neuron SDK helps developers train models on AWS Trainium (and deploy models on the AWS Inferentia chips). It integrates natively with frameworks such as PyTorch and TensorFlow so that you can continue using your existing code and workflows to train models on Trn1 instances.

Amazon EC2 Inf1 instances are purpose-built to deliver high-performance and cost-effective machine learning inference. They provide up to 2.3 times higher throughput and up to 70% lower cost per inference compared to other Amazon EC2 instances. Powered by up to 16 AWS Inferentia chips, ML inference accelerators designed by AWS, Inf1 instances also feature 2nd generation Intel Xeon Scalable processors and offer up to 100 Gbps networking bandwidth to support large-scale ML applications. These instances are ideal for deploying applications such as search engines, recommendation systems, computer vision, speech recognition, natural language processing, personalization, and fraud detection. Developers can deploy their ML models on Inf1 instances using the AWS Neuron SDK, which integrates with popular ML frameworks like TensorFlow, PyTorch, and Apache MXNet, allowing for seamless migration with minimal code changes.

Amazon EC2 G5 instances are the latest generation of NVIDIA GPU-based instances that can be used for a wide range of graphics-intensive and machine-learning use cases. They deliver up to 3x better performance for graphics-intensive applications and machine learning inference and up to 3.3x higher performance for machine learning training compared to Amazon EC2 G4dn instances. Customers can use G5 instances for graphics-intensive applications such as remote workstations, video rendering, and gaming to produce high-fidelity graphics in real time. With G5 instances, machine learning customers get high-performance and cost-efficient infrastructure to train and deploy larger and more sophisticated models for natural language processing, computer vision, and recommender engine use cases. G5 instances deliver up to 3x higher graphics performance and up to 40% better price performance than G4dn instances. They have more ray tracing cores than any other GPU-based EC2 instance.

Amazon EC2 P4d instances deliver high performance for machine learning training and high-performance computing applications in the cloud. Powered by NVIDIA A100 Tensor Core GPUs, they offer industry-leading throughput and low-latency networking, supporting 400 Gbps instance networking. P4d instances provide up to 60% lower cost to train ML models, with an average of 2.5x better performance for deep learning models compared to previous-generation P3 and P3dn instances. Deployed in hyperscale clusters called Amazon EC2 UltraClusters, P4d instances combine high-performance computing, networking, and storage, enabling users to scale from a few to thousands of NVIDIA A100 GPUs based on project needs. Researchers, data scientists, and developers can utilize P4d instances to train ML models for use cases such as natural language processing, object detection and classification, and recommendation engines, as well as to run HPC applications like pharmaceutical discovery and more.

Amazon Elastic Compute Cloud (Amazon EC2) P5 instances, powered by NVIDIA H100 Tensor Core GPUs, and P5e and P5en instances powered by NVIDIA H200 Tensor Core GPUs deliver the highest performance in Amazon EC2 for deep learning and high-performance computing applications. They help you accelerate your time to solution by up to 4x compared to previous-generation GPU-based EC2 instances, and reduce the cost to train ML models by up to 40%. These instances help you iterate on your solutions at a faster pace and get to market more quickly. You can use P5, P5e, and P5en instances for training and deploying increasingly complex large language models and diffusion models powering the most demanding generative artificial intelligence applications. These applications include question-answering, code generation, video and image generation, and speech recognition. You can also use these instances to deploy demanding HPC applications at scale for pharmaceutical discovery.

Amazon EC2 Capacity Blocks for ML enable you to reserve accelerated compute instances in Amazon EC2 UltraClusters for your machine learning workloads. This service supports Amazon EC2 P5en, P5e, P5, and P4d instances, powered by NVIDIA H200, H100, and A100 Tensor Core GPUs, respectively, as well as Trn2 and Trn1 instances powered by AWS Trainium. You can reserve these instances for up to six months in cluster sizes ranging from one to 64 instances (512 GPUs or 1,024 Trainium chips), providing flexibility for various ML workloads. Reservations can be made up to eight weeks in advance. By colocating in Amazon EC2 UltraClusters, Capacity Blocks offer low-latency, high-throughput network connectivity, facilitating efficient distributed training. This setup ensures predictable access to high-performance computing resources, allowing you to plan ML development confidently, run experiments, build prototypes, and accommodate future surges in demand for ML applications.

Amazon EC2 UltraClusters enable you to scale to thousands of GPUs or purpose-built machine learning accelerators, such as AWS Trainium, providing on-demand access to supercomputing-class performance. They democratize supercomputing for ML, generative AI, and high-performance computing developers through a simple pay-as-you-go model without setup or maintenance costs. UltraClusters consist of thousands of accelerated EC2 instances co-located in a given AWS Availability Zone, interconnected using Elastic Fabric Adapter (EFA) networking in a petabit-scale nonblocking network. This architecture offers high-performance networking and access to Amazon FSx for Lustre, a fully managed shared storage built on a high-performance parallel file system, enabling rapid processing of massive datasets with sub-millisecond latencies. EC2 UltraClusters provide scale-out capabilities for distributed ML training and tightly coupled HPC workloads, reducing training times.

Amazon EC2 Trn2 instances, powered by AWS Trainium2 chips, are purpose-built for high-performance deep learning training of generative AI models, including large language models and diffusion models. They offer up to 50% cost-to-train savings over comparable Amazon EC2 instances. Trn2 instances support up to 16 Trainium2 accelerators, providing up to 3 petaflops of FP16/BF16 compute power and 512 GB of high-bandwidth memory. To facilitate efficient data and model parallelism, Trn2 instances feature NeuronLink, a high-speed, nonblocking interconnect, and support up to 1600 Gbps of second-generation Elastic Fabric Adapter (EFAv2) network bandwidth. They are deployed in EC2 UltraClusters, enabling scaling up to 30,000 Trainium2 chips interconnected with a nonblocking petabit-scale network, delivering 6 exaflops of compute performance. The AWS Neuron SDK integrates natively with popular machine learning frameworks like PyTorch and TensorFlow.

AWS Nitro Enclaves enables customers to create isolated compute environments to further protect and securely process highly sensitive data such as personally identifiable information (PII), healthcare, financial, and intellectual property data within their Amazon EC2 instances. Nitro Enclaves uses the same Nitro Hypervisor technology that provides CPU and memory isolation for EC2 instances. Nitro Enclaves helps customers reduce the attack surface area for their most sensitive data processing applications. Enclaves offer an isolated, hardened, and highly constrained environment to host security-critical applications. Nitro Enclaves includes cryptographic attestation for your software, so that you can be sure that only authorized code is running, as well as integration with the AWS Key Management Service so that only your enclaves can access sensitive material.

AWS High Performance Computing (HPC) services empower users to execute large-scale simulations and deep learning workloads in the cloud, providing virtually unlimited compute capacity, high-performance file systems, and high-throughput networking. This suite of services accelerates innovation by offering a broad range of cloud-based tools, including machine learning and analytics, enabling rapid design and testing of new products. Operational efficiency is maximized through on-demand access to compute resources, allowing users to focus on complex problem-solving without the constraints of traditional infrastructure. AWS HPC solutions include Elastic Fabric Adapter (EFA) for low-latency, high-bandwidth networking, AWS Batch for scaling computing jobs, AWS ParallelCluster for simplified cluster deployment, and Amazon FSx for high-performance file systems. These services collectively provide a flexible and scalable environment tailored to diverse HPC workloads.

Elastic Fabric Adapter (EFA) is a network interface for Amazon EC2 instances that enables customers to run applications requiring high levels of inter-node communications at scale on AWS. Its custom-built operating system (OS) bypass hardware interface enhances the performance of inter-instance communications, which is critical to scaling these applications. With EFA, High-Performance Computing (HPC) applications using the Message Passing Interface (MPI) and Machine Learning (ML) applications using NVIDIA Collective Communications Library (NCCL) can scale to thousands of CPUs or GPUs. As a result, you get the application performance of on-premises HPC clusters with the on-demand elasticity and flexibility of the AWS cloud. EFA is available as an optional EC2 networking feature that you can enable on any supported EC2 instance at no additional cost. Plus, it works with the most commonly used interfaces, APIs, and libraries for inter-node communications.