Insights and innovations from AWS re:Invent 2024

Acknowledgment of Contributors
This content was made possible through the insights, expertise, and collaborative efforts of Kent Corley, Alan Henson, Katie Quinn, and Ishaq Zakavi.

AWS re:Invent 2024 delivered another impactful week of groundbreaking announcements, and we’re here to share what it means for your business. From AI and data innovations to tools that simplify global scaling, this year’s event brought game-changing updates designed to help organizations achieve faster time to value, reduce costs, and unlock new growth opportunities.

In this recap, we highlight the most significant announcements and their potential applications, offering insights tailored to help you understand how these innovations can address your challenges and accelerate your goals. Whether you’re exploring generative AI, modernizing legacy systems, or scaling operations globally, the takeaways from re:Invent 2024 directly apply to your journey. Let’s dive in!

Reinventing foundational building blocks and developing brand-new experiences

This wasn’t Matt Garman’s first re:Invent, but it was his first as CEO of Amazon Web Services (AWS). His message was clear: If you aren’t thinking about data and AI, you should be. Garman’s keynote didn’t disappoint, and he kicked off a week of exciting new AWS releases.

His announcements were grouped into five categories: Compute, Storage, Databases, Inference, and Analytics. Let’s hit the highlights of each.

Compute

AWS announced multiple new silicon and EC2 instances to use that silicon. Leading out was AWS Graviton®4 (CPU with ARM architecture), with impressive performance statistics claiming a 30% increase over Graviton3, a 40% increase in database applications, and 45% for large Java applications.

Want a reason to move? Pinterest moved to the Graviton and saw 47% cost savings and 62% carbon emission reductions (thank you, energy-efficient ARM architecture).

A quick nod to the NVIDIA partnership (including Blackwell GPU chip availability), and we were into Trainium™2 (Trn2) announcements, the next-generation AWS GPU silicon. Claiming 30-40% price performance over existing GPU-based instances with up to 20.8 FP8 petaflops with 16 Trn2 chips. Not stopping here, Garman also introduced Trn2 UltraServers. If 16 Trn2 chips didn’t whet your palate, try 4 Trn2 instances connected via NeuronLink™, offering a mind-blowing 83.2 FP8 petaflops.

If you’re considering building one of those Large Language Models, start forming a line now. As a hardware raving fan, the Trn2 UltraServer setup on the exhibition floor at re:Invent was a real highlight.

Storage

No parent claims a favorite, but if forced to choose, Garman made it clear his previous days as the head of S3® product development might bias him a bit. He set the hook by sharing that there are over a million data lakes worldwide, and Parquet is the fastest growing S3 datatype. How else do you respond to this statistic besides offering Amazon S3 Tables, which provides native support to tabular datatypes like Parquet? By also offering Amazon S3 Metadata, a new service that automatically extracts file metadata storing it in S3 Tables. S3 Tables offer 3x faster query performance and up to 10x higher transactions per section for Apache Iceberg tables.

The key takeaway? Removing undifferentiated workloads isn’t just a job for AI. Oh, and traditional databases, you’re on notice.

Databases

Speaking of databases, AWS did what it does and went after the impossible problem using innovative thinking (and no doubt a little historical inspiration). The sun gave us latitude accuracy, but the clock gave us longitudinal accuracy, allowing us to finally know our positioning north-to-south (latitude; see sextant) AND east-to-west (longitude). Time just earned another role – keeping every Amazon Aurora® instance in the world perfectly in sync. If you want to dodge the “Tyranny of the OR” and embrace the “Genius of the AND” to get high availability, multi-region, low latency, strong consistency, zero operational burden, and SQL semantics, you must reduce the hops and handshakes between instances as you guarantee transactional integrity and data propagation. Super accurate and perfectly synced clocks working at microsecond levels mean you can simultaneously send transactions to every distributed instance once. Welcome Amazon Aurora DSQL, the fastest distributed SQL database offering 4x write and read performance. Don’t worry, DynamoDB Global Tables, your time is coming, too (pun intended).

Inference

Welcome to the apex of Garman’s presentation (and an exciting cameo from Andy Jassy, Amazon’s CEO and former AWS CEO). There is far too much here for me to cover. We’re tempted to type keywords and let you have at it on your favorite search engine, but let’s cover the 50,000-foot level.

• Amazon Bedrock^TM Model Distillation: enables knowledge transfer from larger models to smaller models, which can perform up to 500% faster at 75% of the cost
• Amazon Bedrock Knowledge Bases: three words – fully managed RAG (without coding or query management – okay, so that’s eight words)
• Amazon Bedrock Guardrails: generative AI safeguards from detecting sensitive or personal information to grounding responses to a context or guiding models toward identified topics only
• Amazon Bedrock Automated Reasoning checks: leveraging mathematics to verify model response accuracy
• Amazon Bedrock multi-agent collaboration: agentic workflow orchestration without coding
• Amazon Nova^TM (Jassy’s announcements): state-of-the-art foundation models from AWS (text, low-cost multimodal, highly capable multimodal, and advanced multimodal for complex reasoning). There’s also Amazon Nova Canvas (image gen), Amazon Nova Reel (video gen w/ 6 seconds supported now and up to 2 minutes coming), and Amazon Nova Speech-to-Speech and Amazon Nova Any-to-Any coming.
• Amazon Q^TM: now it can generate and apply unit tests, generate accurate code documentation, and perform code reviews. I can’t wait to see this in action. Q is now available in the Console, Slack, VS Code, IntelliJ, GitLab Duo, and others. It can also:
  • Transform .NET applications from Windows to Linux
  • Transform VMWare workloads to cloud-native architectures
  • Transform mainframe application to accelerate migrations (instead, it can help you do this – human-in-the-loop required)
  • Investigate issues across your AWS environment
• Amazon Q Business: Now, you can combine Amazon QuickSight® and Amazon Q Business data. ISVs can now integrate with the Amazon Q Index, and you can automate complex workflows with Amazon Q Business.

Analytics

The most significant announcement was Amazon SageMaker® Unified Studio, a single integrated experience that bridges Analytics, Data, AI, and Catalog & Governance. To accompany this announcement, Amazon SageMaker received a rebranding.

Why is this such a big deal? The pipeline for AI and analytics is convoluted at the organizational and technology levels. Disparate teams work with disparate technologies separated by a common platform (AWS) [see George Bernard Shaw’s “divided by a common language” quote to get the pun]. Bringing the entire workflow together simplifies how the many teams supporting analytics work and work together. From SQL Analytics to Data Processing to Machine Learning to Gen AI Development to Streaming to Business Intelligence to Search Analytics, Amazon SageMaker Unified Studio will have it all in one spot (eventually, stay tuned).

In his keynote, Dr. Sivasubramanian went into great depth about the many offerings from Garman’s Amazon SageMaker Unified Studio announcement. Keep reading for the juicy takeaways.

What does it all mean?

AWS was built with composition in mind. However, you eventually reach a point where your block offerings need composition offerings and AWS delivered. Bringing more capabilities together across compute, storage, databases, inference, and analytics reduces the toil spent on undifferentiated work, making it easier to innovate and build the capabilities that change businesses and communities for the better. AI and data are here to stay, and AWS wants you to have the tools, computing, and storage at the cloud scale to change the world. Happy building!

Data and AI create unique customer experiences

One of the most eagerly anticipated sessions this year was Dr. Swami Sivasubramanian’s AI and Data Keynote. Dr. Sivasubramanian, VP of AI and Data at AWS, did not disappoint with some great new announcements all about empowering AWS’s AI capabilities. You can tell that AWS has been making massive investments to really up its game in these enabling transformational technologies. His theme was all around building on the shoulders of giants.

Amazon Sagemaker Hyperpod provides flexible training plans

If it wasn’t obvious before, it’s now clear that AWS’s focus is to make Sagemaker the single plane of glass to do all your AI development. He started with a strong focus on model training by announcing HyperPod’s flexible training plans. These take away the guesswork from training by allowing for predictable model training timelines and workloads while continuing to benefit from HyperPod capabilities like resiliency, distributed training, and observability and monitoring. These new enhanced HyperPods also allow you to govern your tasks much more easily with centralized governance, optimized and dynamically allocated accelerator utilization across tasks, reducing cost by up to 40 percent and allowing monitoring and auditing compute allocation and utilization in real-time.

As part of the ongoing expansion of the AI ecosystem, many new AI apps are available natively in Sagemaker with no infrastructure to provision or operate. These include launch apps Comet, Fiddler, Deep Checks, and Lakera, with more to come.

Amazon Bedrock announcements highlight robust genAI capabilities

Dr. Sivasubramanian then moved on to more of the inference side with many announcements for Bedrock that continue to build Bedrock out as a robust and flexible GenAI capability.

• Early next year, poolside Assistant, powered by its generative models malibu and point, will be available, which will help address challenges with modern software engineering for large enterprises. These are interesting because they use a new capability called Reinforcement Learning from Code Execution Feedback (RLCEF) to learn from your existing codebase. 
• On the multi-model front, Stable Diffusion 3.5 and Luma AI are coming soon to Bedrock. Luma Ray2, for example, can create a one-minute video. It will be great to see how this works, as it is one of the few products that can produce text-to-video. Bedrock also has a new capability for supporting guardrails for Multimodal Toxicity detection to ensure your GenAI imagery does not go off the rails.
• AWS is taking on Hugging Face® in rolling out Amazon Bedrock Marketplace with access to more than 100 models from leading providers, allowing them to be deployed on custom endpoints using Bedrock APIs for integration and management.
• AWS is also making prompts easier with prompt caching, enabling prompts not to be reprocessed on every call and reducing costs by 90% and latency by up to 85%. Bedrock will also be able to route prompts to the best model using intelligent prompt routing. Kendra is expanding through the Kendra Generative AI index, which supports connectors to 40+ enterprise data sources to integrate with Bedrock knowledge bases and Amazon Q business.
• RAG and data tools are becoming much more powerful with new capabilities allowing Bedrock Knowledgebases to support structured data retrieval and GraphRAG so you can use Graph Databases in Bedrock knowledgebases. Data Automation is expanding with GenAI ETL (or no ETL, as Dr. Sivasubramanian put it), allowing you to help transform unstructured data, such as videos or documents, into structured formats ready for analysis.
• Continuing to further expand Sagemaker as the AI developer’s tool of choice, Amazon Q Developer^TM is now available in Sagemaker, allowing you to develop machine learning models using natural language “without a single line of Python code.” As time passes, we will see how realistic this vision is, as you always have to get back to the code at some point. Amazon Q is also being expanded in Amazon QuickSight Scenarios to help users break down complex queries into manageable steps, enabling faster decision-making.

AWS education literacy initiative

Finally, Dr. Sivasubramanian highlighted the AWS education literacy initiative, which provides a five-year commitment to bringing cloud technology and technical support to organizations creating digital learning solutions. This commitment reflects an investment by AWS of 100 million dollars in cloud credits and support over the next five years.

There were many significant announcements in training, inference, RAG, data, and Q to further enhance AWS’s impressive portfolio of AI capabilities. Our teams are excited to deploy, test, and learn more about these great new features with our clients and partners. 

Dr. Werner Vogels, VP and CTO at Amazon.com, addressed the fundamental challenge that scaling global systems is hard and complexity is here to stay. But rather than hiding it, he introduced the idea of “simplexity.”

This approach embraces complexity under the surface—like globally distributed architectures—while ensuring the end-user experience remains straightforward. The goal isn’t to pretend things are simple but to engineer solutions so effectively that complexity never spills over to customers.

Complexity and architecture

Vogels emphasized that complexity isn’t a problem to eliminate but to manage. Modern architectures—think microservices, cell-based patterns, and designs that prioritize predictability—can handle growth and evolve naturally. By aligning team structures with technical architectures, organizations can ensure the right people are positioned to adapt their systems as they expand.

Foundational innovations: Amazon Aurora DSQL, Clock Bound Daemon and more

One standout announcement was Amazon Aurora DSQL: a multi-region, multi-active database designed for truly global operations. Built on precise clock synchronization and robust replication, it sets a new standard for consistency and availability. Paired with the Amazon Time Sync Service enhancements, developers get more confident reasoning about distributed data. The open-source Clock Bound Daemon offers a reliable “sense of time” for distributed events—something traditionally tough to nail down.

Sustaining simplicity through culture and automation

Technology alone isn’t enough. Vogels highlighted the importance of team culture—embracing continuous improvement, shared ownership, and a healthy respect for potential failures. Automation plays a key role in offloading repetitive tasks so engineers can focus on strategic improvements. Whether it’s employing Clock Bound Daemon or leveraging Agential Workflows, this combination of culture and tooling ensures complexity doesn’t become the customer’s problem.

Real-world confidence

This isn’t just conceptual. Vogels showcased customers who started small and scaled globally without compromising user experience. Techniques like iterative decomposition, strategic caching, and selective automation prove that complexity can be managed and tamed. It’s a practical path, not just a theoretical ideal.

What does it all mean?

As organizations expand their global reach and add new features—especially those powered by AI—complexity will only increase. Vogels’ message is reassuring: with the right building blocks (like Aurora DSQL and precise time services), a deliberate culture, and intelligent automation, you can keep complexity where it belongs—behind the scenes.

Ultimately, Vogels differentiated between intentional complexity (the kind that enables better capabilities) and the unintended complexity that comes from quick fixes and legacy code.

“Simplexity” encourages us to plan for growth from day one, break down complexity into manageable pieces, align people with technology, and automate wherever possible. It’s a reminder that, done right, complexity doesn’t have to get in the way

AWS and Pariveda create a positive impact for customers

Dr. Ruba Borno, Vice President, Global Specialists and Partners, emphasized in the Partner Keynote what we at Pariveda have already experienced firsthand: AWS’s collaboration with its Partners like Pariveda is crucial for developing innovative solutions to speed up time to value.

Key engagement strategies were outlined, including using AWS Marketplace solutions, AWS funding opportunities that partners can leverage through the Migration Assistance Program (MAP), and AWS’s commitment to collaborating with Partners.

New releases to reduce costs and increase efficiency

• Gen AI tooling like AWS’s Q Developer will help organizations move even faster. This is especially important in the context of modernization projects that are often more expensive than anyone desires but are necessary to maintain applications driving key operations.

• AWS is releasing additional Q Developer Transform tooling to help with .NET, Java, and legacy tooling modernizations and upgrades. Productivity gains are forecasted to be up to 30% on standard tasks, with double the gain on modernization tasks to help decrease overall cost on maintenance and migration solutions.

• AWS is committed to helping further quick-start AWS migrations and innovative solutions by expanding the Migration Assistance Program (MAP), which will not have a monetary cap on the available amount. This allows for less stress on budget constraints, and Pariveda actively works with organizations to identify MAP funding to assist in the process of considering average yearly projected cloud costs.

• Looking ahead, AWS and Pariveda are both focused on expanding their global presence this year.

AWS Marketplace solutions create quicker time-to-value

AWS and Pariveda work together to help organizations solve unique challenges. Unsurprisingly, 80% of greenfield migrations involve a partner, underscoring the critical role partners like Pariveda play in successful cloud transformations.

At Pariveda, we have worked to develop various AWS Marketplace solutions such as Cloud Cost Optimization Assessment, Data Engineering Accelerator, and a Gen AI POC Accelerator, which are trusted approaches that can be leveraged for quicker time to value.

Dr. Borno shared that ProServe continues to be an important lever organizations can use to ensure the quality of their solutions.

Jane Lacy, AWS Director of Global Public Sector SI Partner Sales, highlighted that Pariveda is the only partner out of all her 1500 partners who have figured out a positive, productive working model with ProServe, which is a key differentiator.

Empowering innovation and transformation with AWS and Pariveda

AWS re:Invent 2024 showcased a clear vision for the future—one that empowers businesses to innovate, modernize, and scale like never before. From cutting-edge advancements in generative AI and data tools to programs designed to simplify migrations and reduce costs, AWS continues to deliver solutions that address the most pressing challenges facing organizations today.

The opportunities for our clients are immense. Whether you’re leveraging AWS Marketplace solutions to accelerate time-to-value, exploring the transformative power of generative AI, or tapping into programs like MAP to drive cost-effective modernization, the tools are now at your fingertips.

At Pariveda, we are committed to guiding you through these innovations, ensuring they align with your unique goals and unlock meaningful impact for your business. As you plan your next steps, remember: the journey to transformation doesn’t have to be taken alone. With AWS’s robust ecosystem and Pariveda’s expertise, you can navigate complexity, embrace innovation, and create a future that’s not just scalable but extraordinary. Let’s build that future together.