- WeeklyCloud.info
- Posts
- Weekly Cloud Info #W29 - 2025
Weekly Cloud Info #W29 - 2025
Walid LARABI
July 18, 2025
Hi!
This week: Amazon EKS scales to 100K nodes for massive AI/ML workloads. AWS debuts native vector storage and Bedrock AgentCore for scalable AI agents. GKE and AKS boost resilience with smarter backups and auto-scaling.
Have a great read. ☕
📰 Top picks of the week
Amazon EKS Expands Support to 100,000 Worker Nodes for Advanced AI/ML Workloads
Amazon Elastic Kubernetes Service (EKS) now supports up to 100,000 worker nodes in a single cluster, allowing customers to scale to 1.6 million AWS Trainium accelerators or 800,000 NVIDIA GPUs for large AI/ML models. This enhancement enables efficient coordination of extensive AI workloads, reduces operational costs, and provides flexibility in using various AI/ML frameworks while maintaining Kubernetes compatibility.
AWS Introduces Model Context Protocol Server for Enhanced S3 Tables Management
AWS has launched the Model Context Protocol (MCP) Server for Amazon S3 Tables, enhancing AI-assisted data management. This integration allows AI code assistants to understand and interact with S3 Tables using natural language for tasks like table creation, schema definition, and data operations. Developers can install the MCP Server and configure their AWS accounts to leverage these capabilities.
AWS Launches S3 Vectors for Cost-Effective Vector Storage and Querying
Amazon has announced the preview of Amazon S3 Vectors, a new storage solution designed for large vector datasets, capable of reducing costs by up to 90%. It features vector buckets with dedicated APIs for storing and querying vector data, supporting up to 10,000 vector indexes and millions of vectors. S3 Vectors integrates with Amazon Bedrock and OpenSearch Service, facilitating scalable generative AI applications and efficient data management without the need for complex infrastructure.
AWS Launches Amazon Bedrock AgentCore for Scalable AI Agent Management
Amazon has introduced the preview of Amazon Bedrock AgentCore, a suite of services designed to help developers deploy and manage AI agents at scale. Key features include AgentCore Runtime for low-latency environments, AgentCore Memory for managing session and long-term memory, AgentCore Observability for visualizing agent execution, AgentCore Identity for secure access to AWS and third-party services, AgentCore Gateway for transforming APIs into agent-ready tools, and AgentCore Browser for web automation workflows.
Amazon S3 Metadata Enhances Object Visibility with Live Inventory Tables
Amazon S3 Metadata now offers complete visibility into all objects in S3 buckets, allowing users to query metadata for existing objects through live inventory tables. These tables provide a near real-time view of object changes and are automatically updated within an hour of modifications. This feature simplifies metadata management, supports analytics, and enhances auditing capabilities without the need for custom systems.
AKS Introduces General Availability of Virtual Machines Node Pools
Azure Kubernetes Services (AKS) now offers generally available support for Virtual Machines node pools, allowing users to manage multiple VM SKUs of a similar family within a single node pool. This feature simplifies workload deployment by reducing the need for separate node pools for each SKU type, streamlining node management.
Azure AKS Introduces Node Auto-Provisioning for Enhanced Resource Management
Azure Kubernetes Service (AKS) now supports Node Auto-Provisioning (NAP), which automatically provisions single-instance nodes in response to unscheduled pods. This feature eliminates the need for pre-configured node pools, allowing for on-demand scaling that enhances resource efficiency, simplifies cluster management, and improves cost control for dynamic workloads.
Microsoft Fabric Introduces Public Preview of Cosmos DB Integration for Enhanced Data Workflows
Microsoft has launched the open Public Preview of Cosmos DB in Microsoft Fabric, allowing users to access and analyze operational data seamlessly within the Fabric ecosystem. This integration combines Azure Cosmos DB's capabilities with real-time intelligence and AI-optimized performance, enabling users to store both NoSQL and relational data. Key features include automatic mirroring of operational data to OneLake for real-time analytics and support for SQL queries, enhancing data workflows across applications.
Google Cloud Introduces Cross-Project Backup for GKE to Enhance Data Protection and Disaster Recovery
Google Cloud's Backup for GKE, now in preview, introduces cross-project backup and restore capabilities. This feature allows users to back up workloads from one GKE cluster, store them in a separate project, and restore them in another, enhancing data protection and disaster recovery. It simplifies centralized backup management, boosts resilience against regional outages, and streamlines operations by enabling the seeding and cloning of environments across projects.
Google Cloud Launches Easy Bucket Relocation Feature for Cloud Storage
Google Cloud has introduced a new feature called Cloud Storage bucket relocation, allowing users to easily change the location of their storage buckets without complex planning or extended downtime. This feature preserves bucket names and object metadata, ensuring minimal application disruption and maintaining original storage classes. It utilizes asynchronous data copying and metadata preservation techniques to facilitate smooth transitions while optimizing storage costs.
❤️ You might also like
🎁 This week hidden gem
Running July 1–October 31, participants can access no-cost Oracle Cloud training and earn certification badges for various tracks during this challenge https://education.oracle.com/race-to-certification-2025
🏁 Enjoy this newsletter?
Forward it to a friend, and let them know they can subscribe here.