AWS Reinvent 2023 Recap

Published in

Lumigo

10 min readDec 5, 2023

This year’s conference marked a significant milestone in its history, characterized by an atmosphere of anticipation and excitement that each reinvent brings. As the first keynote got underway, all eyes were fixed on the stage, where many speakers unveiled many new features, products and announcements. This is the moment that many have been eagerly awaiting the announcements that help shape cloud ecosystems for the years to come.

At Reinvent, our booth, featuring 1-click OpenTelemetry capabilities and engaging live demonstrations, became a hub of excitement. Attendees were particularly drawn to our innovative approach that enables tracing applications on Amazon EKS, ECS, and Lambda without requiring code modifications. The ease of integrating this no-code solution stood out, especially with our implementation of a Kubernetes operator that automatically handles the instrumentation needed for effective tracing within a specific namespace.

Many were particularly impressed by how our solution prioritizes ease and efficiency in observability deployments and empowers developers with insightful trace maps of their applications. This feature enables straightforward navigation through the plethora of services and intricacies inherent in their deployed environments. Such versatility directly responds to an evolving trend in the industry, where organizations are increasingly blending a variety of AWS services. This approach aims to achieve both flexibility and optimal performance, especially in environments that mix serverless and container-based architectures.

Thinking outside the EKS pod

One of the most important announcements at Reinvent regarding Amazon Elastic Kubernetes Service (EKS) is the introduction of Amazon EKS Pod Identity.

This new feature dramatically simplifies the way applications within EKS access AWS services. It provides an efficient and straightforward method for defining the required IAM permissions for applications in EKS clusters. With EKS Pod Identity, managing permissions across multiple EKS clusters becomes more manageable, offering a seamless and user-friendly experience in securing access to AWS services.

The integration of EKS Pod Identity streamlines the process of IAM role management for Kubernetes applications. To use this feature, you start by creating an IAM role with the necessary permissions, specifying pods.eks.amazonaws.com it as the service principal in its trust policy. Following this, the Amazon EKS Pod Identity Agent add-on is installed via the Amazon EKS console or the AWS CLI. The final step involves mapping the IAM role to a service account, which can be done directly in the EKS console, through APIs, or using the AWS CLI. This setup ensures that new pods associated with the service account are automatically configured with the appropriate IAM credentials, simplifying the deployment and scaling of applications in EKS.

Moreover, Amazon EKS Pod Identity enhances flexibility through new AWS APIs, catering to advanced and automated configurations. Developers and administrators can leverage these APIs to install the EKS Pod Identity Agent add-on or map IAM roles to Kubernetes pods using AWS CLI commands like aws eks create-addon and aws eks create-pod-identity-association. These functionalities underscore AWS’s dedication to improving the EKS experience, focusing on security, efficiency, and ease of management in cloud-native environments.

Reinvent — Day 0

Kicking off Reinvent 2023, AWS made a splash with major announcements, particularly in the serverless landscape. A standout update is Lambda’s enhanced scaling, which is now 12 times faster, allowing functions to burst up to 1000 concurrent executions instantly. This improvement is crucial for handling sudden traffic spikes, potentially supporting up to 10,000 requests per second per endpoint. Additionally, Step Functions received a significant update, enabling direct calls to public HTTP endpoints, thereby simplifying interactions with third-party APIs. Another critical addition to Step Functions is the capability to test individual states, streamlining the debugging process for complex workflows.

CloudWatch introduced an AI assistant for querying logs and metrics in plain English, marking a step forward in user-friendly cloud monitoring. Alongside this, CloudWatch’s new Infrequent Access log class offers a cost-effective solution for log storage, complemented by introducing anomaly detection features. EventBridge expanded its integration capabilities, notably with Adobe and Stripe, enhancing its event-driven architecture support. This move from webhook-based solutions to fully managed integrations signifies a leap in ease of use and efficiency.

In infrastructure management, AWS CloudFormation’s new synchronization with Git templates simplifies deployment processes, which is particularly beneficial for smaller operational stacks. This feature offers an alternative to traditional CI/CD pipelines for specific use cases. Meanwhile, Step Functions’ optimized integration with Bedrock significantly boosts AI application development, although Lambda and Function URLs remain crucial for scenarios needing response streaming. Day 0 of Reinvent 2023 thus set a high standard for innovation, especially in serverless solutions and cloud infrastructure management.

Reinvent — Day 1

Day 1 at Reinvent 2023 presented various updates, with some exciting developments and a notable letdown. One of the major talking points was the launch of Elasticache “serverless”, which, akin to OpenSearch “serverless”, fell short of expectations. Based on data storage and processing unit, the pricing model raised eyebrows due to its minimum metered storage cost, amounting to a significant $90 per month even with no data.

AWS also unveiled Aurora Limitless Database, promising to scale Amazon Aurora clusters to manage millions of write transactions per second and handle petabytes of data. While such extreme scalability might be overkill for most, the technical prowess behind it is undeniably impressive. In the realm of messaging, SQS FIFO throughput received a major boost, and it is now capable of processing up to 70,000 messages per second in high throughput mode. Additionally, AWS has extended the dead-letter queue (DLQ) redrive support to FIFO queues, a significant improvement that allows for the redriving of a FIFO DLQ to another FIFO queue, enhancing message processing reliability.

Another significant announcement was the integration of Step Functions with AWS App Composer. This collaboration marks a leap forward in workflow management, enabling users to seamlessly incorporate state machines into their stacks and edit them directly in the Step Functions Workflow Studio. This integration signifies AWS’s ongoing commitment to streamlining and enhancing the user experience in application development and workflow orchestration.

Reinvent — Day 2

The second day of Reinvent 2023 was headlined by Adam Selipsky’s keynote, which focused heavily on AI developments. A major highlight was the introduction of Amazon Q, AWS’s response to ChatGPT. Integrated within the AWS console, documentation pages, and even in your IDE through the VS Code plugin, AWS Toolkit, Amazon Q is trained on AWS documentation, potentially offering more current answers than ChatGPT. However, early testing suggests that while it’s adept at providing factual information, its advice on best practices may not be fully reliable yet. What sets Amazon Q apart is its data security aspect, allowing users to grant it access to specific AWS environment data, and its coding assistance capabilities, including Java code transformation and feature implementation with CodeCatalyst.

Day 2 also saw significant updates to Bedrock, including introducing guardrails to redact PII data, filter harmful content, and block certain topics. The Bedrock Knowledge Base, akin to OpenAI’s GPT models, is now generally available, enabling the creation of customized chatbots using managed Retrieval Augmented Generation (RAG). It simplifies the ingestion of data into a vector database, although the cost associated with OpenSearch Serverless vector store may lead users to consider alternative databases. Another noteworthy release is Bedrock Agents, enhancing the development of autonomous AI agents.

In addition to AI advancements, AWS announced the new S3 Express One Zone storage class. This storage class offers single-digit millisecond latency and maintains the same high durability, storing data in a single Availability Zone. While it provides a cost advantage in terms of request pricing, its storage cost is higher, attributed to the use of specialized hardware that integrates storage with compute for improved performance.

AWS’s announcement of new no-code integrations, particularly the DynamoDB to OpenSearch integration, sparked interest. However, the lack of OpenSearch as an EventBridge Pipes target and incomplete CloudFormation support for these integrations was a point of disappointment. Users seeking Infrastructure as Code (IaC) solutions may need to resort to using the DynamoDB console for initial setup, extracting the YAML format for OpenSearch ingestion from the wizard process. Day 2 of Reinvent 2023 thus mixed exciting AI and storage innovations with certain aspects that left room for improvement, reflecting AWS’s continuous push towards simplifying and enhancing cloud technology.

Reinvent — Day 3

Day 3 of Reinvent 2023 continued the AI-centric theme, showcasing new models for Bedrock and advancements in vector search capabilities. Bedrock introduced several foundational models, including Anthropic Claude 2.1, Amazon Titan Text models, Titan Multimodal Embeddings, Titan Image Generator model, Stable Diffusion XL 1.0, and Meta Llama 2 70B. These arrivals are timely, especially considering the recent vulnerabilities found in ChatGPT and changes in its coding abilities. Additionally, Bedrock now supports batch inference, a feature whose applications could range from background tasks by autonomous agents to grouping user prompts in high-throughput AI applications, thereby reducing the number of API requests.

SageMaker’s new HyperPod feature promises to reduce the time required for training large language models (LLMs) by 40%, a significant advancement for AI development. This could potentially halve the time for model iterations, which is really important given that training LLMs is usually time-intensive. AWS announced vector support for various databases in the vector search domain, including a vector engine for OpenSearch Serverless, vector search for Amazon MemoryDB for Redis, and Amazon DocumentDB. The expectation of DynamoDB receiving vector search capabilities led to a moment of excitement. Still, it turns out that vector search for DynamoDB data requires routing the data through OpenSearch first, as per the no-code integration introduced on day 2.

OpenSearch received considerable focus, with the introduction of a new instance family, OR1, boasting a 30% improvement in price-to-performance ratio. Neptune, AWS’s graph database service, unveiled a new analytics engine, expanding its capabilities for graph analytics on data within Neptune databases or S3. This enhancement opens up new possibilities for use cases like fraud detection, game economy modelling, and complex data analysis. Amidst the AI-focused announcements, introducing a new API for AWS Marketplace was somewhat unexpected. This API aims to streamline orchestration for SaaS companies selling products on AWS Marketplace, potentially enhancing the end-user experience.

Reinvent — Day 4

Day 4 of Reinvent 2023 was marked by Werner Vogel’s keynote, always a much-anticipated event at the conference. While there were no major service announcements, Werner emphasized the importance of considering cost as a critical non-functional requirement in architectural design. He advocated for the role of the “frugal architect,” underscoring the need for cost-efficiency in cloud architecture. This message resonates strongly in today’s environment, where optimizing resources and managing expenses are key to sustainable and efficient cloud operations.

The day also saw significant enhancements in development tools, notably with App Composer’s arrival in VS Code. This integration allows developers to visually design and manage their Serverless Application Model (SAM) or CloudFormation templates directly within VS Code, a boon for those working extensively with these AWS services. CloudWatch Application Signals was also introduced, offering automated instrumentation for Java applications running on ECS, EKS, and EC2. This development promises to simplify monitoring and troubleshooting, although its effectiveness remains to be seen.

Amazon Inspector received an update, now capable of continuous monitoring of EC2 instances and employing genAI to analyze and flag security vulnerabilities in Lambda functions. While these capabilities are impressive, they come with a potential cost of up to $0.90 per function per month, which could be a surprise expense for many users. However, AWS offers other AI-assisted tools like Amazon Q, Copilot, and CodeWhisperer, which can provide similar functionalities without these additional costs. The advantage of Amazon Inspector lies in its automated nature and the ability to be activated across an entire AWS Organization from a delegated account, aiding central security teams in providing governance and support.

Reinvent concludes for another year

Reinvent, AWS’s flagship event, stands as a testament to the power of cloud technology in bringing together a diverse and global community of developers, architects and DevOps. Each year, this conference showcases the latest advancements in cloud computing and serves as a melting pot of ideas and innovations. The event highlights AWS’s commitment to fostering a rich, inclusive, and ever-expanding cloud community, underscoring the idea that cloud technology is a unifying force in the modern digital landscape.

Reinvent’s significance extends beyond its technological showcases; it’s a celebration of the cloud community’s collective ingenuity and vision. The diverse range of participants — from startups to large enterprises — reflects the universal appeal and applicability of cloud solutions. Through keynotes, workshops, and discussions, Reinvent catalyzes cross-border and cross-industry collaborations, setting the stage for the next wave of cloud innovations. It’s a unique opportunity for participants from all corners of the globe to engage, inspire, and be inspired, reinforcing the notion that a few do not just shape the future of technology but a global community united by the cloud and a love of building scalable solutions.

If you enjoyed reading this post, please clap and also share; it goes a long way to helping support the effort we put into creating such content. Additionally, if you enjoyed it, then please share so others can as well!