Ready for the latest cloud innovations from AWS?

Here's what you need to know from re:Invent 2024!

What is AWS re:Invent?

AWS re:Invent is Amazon's flagship cloud event, running from December 2–6 in Las Vegas, Nevada. It’s the ultimate showcase for cloud innovations, featuring keynotes, technical sessions, and announcements of groundbreaking technologies. While in-person tickets cost $2,000, you can register to watch the event online for free.

Check out our latest video for more info and a visual breakdown of these announcements:

This Week’s Overview

Throughout the week, I’ll be summarising the key takeaways from each day’s sessions. If you prefer high-level video content, check out our YouTube channel for daily recaps under 10 minutes.

This year’s keynotes and talks will cover major advancements, from AI updates like Trainium2 chips and Bedrock enhancements to innovations in networking, storage, and custom silicon. With over 18 Innovation Talks, 5 Keynotes, and countless sessions, there’s plenty to look forward to—whether you’re interested in AI, analytics, or cutting-edge cloud technology.

Stay tuned for daily updates in your inbox and on YouTube, where I’ll break down the most exciting announcements and insights from AWS re:Invent 2024!

Day 1

Key Takeaways

• AWS unveiled Graviton4, their latest custom silicon designed for large-scale workloads.

• The AWS Nitro System saw significant improvements, enhancing both security and performance.

• Disaggregated storage architecture introduced for improved scalability.

• New AI capabilities announced with Trainium2 chips and the Ultra Server.

• 10u10p networking fabric revealed for handling AI’s demanding needs.

Full Summary

Peter DeSantis, AWS Senior VP of Utility Computing, showcased key innovations during the 2024 re:Invent keynote, including advancements in custom silicon (Graviton4, Trainium2), the Nitro System, disaggregated storage architecture, and AI-optimised technologies like Bedrock updates and 10u10p.

Emphasising AWS's culture of detail and operational rigour, the presentation highlighted AWS's commitment to driving performance, scalability, and innovation across cloud and AI workloads.

Below you can find the highlights and timestamps for the keynote in case you want more details.

Utility Computing and Technical Innovation

• Peter DeSantis, Senior VP of Utility Computing at AWS, kicked off the 2024 re:Invent keynote with a deep dive into AWS's core technical innovations. (00:00:21)

• He highlighted how AWS differentiates itself through significant investments in cloud infrastructure. (00:00:52)

• Using the analogy of trees, DeSantis explained AWS's focus on innovation, emphasizing that just as a tree's roots support growth, AWS’s attention to detail supports its cloud services. (00:03:16)

• Key AWS innovations showcased include investments in custom silicon, AWS hypervisors, and database technologies, designed to meet the growing demands of cloud computing. (00:04:22)

Graviton4 and Nitro: Advances in Custom Silicon and Security

• AWS highlighted its continued commitment to custom silicon with the announcement of Graviton4. (00:11:10)

• Graviton4 represents a significant improvement over previous versions, supporting large-scale workloads with multi-socket capabilities and improved compute performance. (00:12:32)

• Graviton4 is particularly aimed at demanding enterprise use cases. (00:13:40)

• The AWS Nitro System was also a focal point, demonstrating how AWS has reimagined server architecture to improve both security and performance. (00:18:09)

• Nitro’s enhanced security capabilities include hardware supply chain integrity and cryptographic attestation to ensure end-to-end trust. (00:19:00)

Disaggregated Storage Architecture

• DeSantis unveiled AWS’s new disaggregated storage architecture. (00:27:28)

• The new system uses AWS Nitro cards embedded in JBOD ("just a bunch of disks") enclosures to separate compute from storage. (00:30:33)

• This disaggregated approach makes AWS’s storage infrastructure more efficient, scalable, and easier to maintain. (00:33:59)

• It provides better performance for services like Amazon S3, EBS, and EFS. (00:34:19)

Introducing Trainium2 and Ultra Server for AI

• AWS introduced new Trainium2 chips and Ultra Server, focusing on artificial intelligence workloads. (00:46:41)

• The Trainium2-powered Ultra Server is AWS's most powerful AI server yet, featuring 64 Trainium2 chips and 1.5TB of high-speed HBM memory. (00:47:48)

• The unique system architecture is optimized for AI workloads. (00:47:58)

• Trainium2’s systolic array architecture is specifically designed to handle large-scale AI models efficiently. (00:56:04)

• It leverages specialized hardware and memory access patterns to reduce latency and optimize compute. (00:56:46)

AI Inference Optimization and Bedrock Updates

• AWS announced a new latency-optimized option for Amazon Bedrock. (01:04:27)

• The update is designed to improve inference performance for AI models like LLAMA and Anthropic’s Claude 3.5. (01:04:56)

• AI models can now leverage Trainium2, offering significantly reduced latency for inference workloads. (01:05:15)

• This benefits customers who need fast, cost-effective AI capabilities. (01:05:41)

10u10p Network Fabric for Scalability

• AWS introduced its latest AI-optimized networking fabric, 10u10p, to handle the demanding requirements of AI training and inference. (01:13:04)

• The network fabric delivers massive bandwidth and low latency, scaling effectively to support thousands of servers. (01:13:18)

• AWS also unveiled proprietary trunk connectors and Firefly optic plugs. (01:17:20)

• These innovations improve network reliability and simplify setup, leading to faster deployment and fewer issues in data centres. (01:18:41)

AWS’s Unique Culture of Detail

• DeSantis emphasized AWS's culture of detail, operational rigour, and commitment to innovation across all layers of infrastructure—from chips to network fabrics. (01:12:44)

• This culture enables AWS to respond to customer needs quickly and stay ahead of the curve with game-changing technologies for cloud and AI workloads. (01:12:57)

Stay Tuned

Check back tomorrow for more insights from AWS re:Invent 2024. For full coverage, make sure to visit our YouTube channel for video highlights and in-depth discussions!

Luke