AWS re:Invent 2024 - Compute
Back to all Re:Invent 2024
Table of Contents
AI & ML
 |
AWS introduces the powerful Trainium2 chip, designed to provide breakthrough AI training and inference performance. The talk highlights the chip's advanced features, including high-performance compute, cost-efficiency, scalability, and innovative capabilities that enable the next generation of frontier models.
|
 |
The video discusses the latest advancements in AWS AI chips, including the Trainium2 and UltraServer, which provide unprecedented performance and scale for large language models and other AI applications. The video also highlights the portability and usability of the Neuron SDK, including the integration with JAX, and showcases the performance and optimization capabilities of the Neuron platform.
|
 |
The video discusses the latest trends in generative AI, including the use of large language models, small language models, multi-modal models, and agentic AI. It also highlights how companies like Ricoh, Arcee AI, IBM, and ByteDance are leveraging AWS AI chips Trainium and Inferentia to optimize the performance and cost of their generative AI workloads.
|
 |
The video discusses how AWS accelerated computing enables customer success with generative AI, including use cases across industries, trends in LLM training and inference, and the capabilities AWS is developing in EC2 to support performance, cost, security, and ease of use. It also features a case study from Meta on how they used AWS to build a multimodal AI model for their smart glasses.
|
Compute
 |
This talk provides an overview of AWS Graviton, the latest generation of custom ARM-based processors from AWS, and how they deliver the best price-performance for a variety of workloads. The speakers discuss the technical details of Graviton4, customer adoption stories, and best practices for transitioning to Graviton-based instances.
|
 |
The video covers various capacity management models offered by AWS EC2, including on-demand, on-demand capacity reservations, Capacity Blocks, and Spot instances. It highlights the key features, benefits, and use cases of each model, as well as the tools and techniques available to customers for optimizing capacity and cost efficiency.
|
 |
The session explores how leading semiconductor companies are leveraging AWS to overcome the evolving challenges in the EDA industry, such as increasing design complexity, time-to-market pressures, and the need for global collaboration and access to the latest technologies. Astera Labs, a successful semiconductor startup, shares its journey of running 100% of its EDA workloads on AWS, highlighting the benefits of scalability, elasticity, and cost optimization.
|
 |
The presentation covers the AWS EC2 High Memory portfolio for running SAP HANA workloads, highlighting the benefits of the latest U7inh instances with 32TB of memory and 1,920 vCPUs. The speakers also discuss the technical underpinnings of the Nitro system, the performance and storage capabilities, and how SAP leverages these instances for its own and customer workloads.
|
 |
The AWS Graviton Savings Dashboard provides comprehensive insights into a company's Graviton processor usage and identifies significant cost-saving opportunities. The dashboard's interactive visualizations and detailed analysis empower FinOps teams and engineers to make data-driven decisions and accelerate their Graviton adoption journey.
|
 |
The talk explores the evolution of large language models and the AWS infrastructure required to train them at scale. It highlights the services and tools available on AWS to provision, manage, and optimize distributed training workloads for high-performance and resilience.
|
 |
The video discusses how AWS enables innovation and results with high-performance computing (HPC) services, showcasing customer journeys of Merck and PhysicsX. It also highlights the convergence of HPC and AI, particularly in weather and climate modeling, and the strong partner ecosystem that supports these advancements.
|
 |
This talk discusses the current state of quantum computing, the potential applications, and the collaboration between AWS and NVIDIA to accelerate R&D in this field. It highlights the challenges in building a production-ready quantum computing ecosystem and the efforts to integrate quantum and classical computing to enable new use cases.
|
 |
The presentation covers the latest updates and roadmap for Amazon Linux, including the extension of support for Amazon Linux 2 and the introduction of new features and packages in Amazon Linux 2023. The speakers also discuss the challenges and best practices for migrating to the new version, particularly around compliance and security requirements.
|
 |
This session provides an in-depth overview of AWS Confidential Computing, including the Nitro System and Nitro Enclaves, and how customers like 1Password are leveraging these technologies to protect sensitive data while in use. The presenters explain the key features, benefits, and use cases of confidential computing, emphasizing the importance of understanding what data needs protection and from whom.
|
 |
This session explores strategies and techniques to maximize Amazon EC2 savings and improve performance, including the use of Graviton instances, Spot Instances, and savings plans. The speakers discuss the benefits of these approaches, share insights from Nubank's journey to cloud efficiency, and provide guidance on overcoming common challenges.
|
 |
The AWS Nitro System is a fundamental rethink of virtualization in the cloud, providing better performance, security, and innovation through specialized hardware and software. The talk covers how the Nitro System offloads networking, storage, and security functionality from the host CPU to dedicated chips, resulting in bare-metal-like performance and enhanced security features.
|
 |
The talk covers the evolution of Amazon EC2 over the years, including the introduction of new instance types, processors, and networking capabilities. It also highlights AWS's efforts to help customers make informed decisions about their cloud infrastructure and optimize their usage through tools like the AWS Savings Dashboard and the EC2 Instance Finder.
|
 |
This session discusses how Monzo, a digital bank, optimized their workloads on Amazon EKS using Karpenter and EC2 Spot Instances. The presentation covers Monzo's journey from self-managed Kubernetes to leveraging EKS, Karpenter, and Spot Instances to achieve significant cost savings while maintaining application scalability and resilience.
|
Developer Experience
 |
The video discusses how Block, a financial services company, migrated their Apple platform development infrastructure from on-premises to AWS EC2 Mac instances. The key highlights include the challenges faced with the on-premises setup, the benefits of adopting EC2 Mac, and the technical details of how the migration was implemented and the lessons learned.
|
HPC
 |
The video discusses how AWS is reinventing high-performance computing (HPC) to help users think big. It covers AWS's Parallel Computing Service, the company's strategy for supporting HPC workloads, and customer success stories in areas like renewable energy, fusion power, and gene sequencing.
|
Quantum Computing
 |
The presentation discusses AWS's quantum computing efforts, including the launch of the Amazon Braket service, the company's own quantum hardware development, and the journey of enterprise customers in adopting quantum computing. The key focus is on the new Quantum Embark program, designed to help enterprises at the early stages of their quantum computing exploration.
|