AWS

AWS Glue Schema Registry is now available in ten additional regions, including Asia Pacific, Mexico, Israel, and Canada West. This serverless feature allows teams to validate and control the evolution of streaming data using Apache Avro, JSON, and Protobuf formats. It serves as a centralized repository to manage data structure between decoupled applications in streaming systems.

• Schema Registry is now live in 10 new regions for global data governance.
• Supports Apache Avro, JSON, and Protobuf for streaming data validation.
• Eliminates custom data validation logic and reduces cross-team coordination.
• Improves data quality and prevents downstream application failures.

Amazon OpenSearch Service now allows administrators to restrict access to OpenSearch UI applications using standard IAM condition keys like aws:SourceVpc and aws:SourceIp. This feature enables a consistent data perimeter by supporting network restrictions at the identity-based, VPC endpoint, and resource control policy levels. It integrates seamlessly with existing AWS security configurations to limit access to approved networks.

• Restrict OpenSearch UI access using aws:SourceVpc, aws:SourceVpce, and aws:SourceIp IAM keys.
• Enforce network policies at three levels: identity, VPC endpoint, and resource control policies.
• Maintain a consistent data perimeter across AWS environments with unified access controls.
• No new IAM keys required; leverage existing standard conditions for network restrictions.

AWS Backup now supports logically air-gapped vaults for Amazon Neptune in Asia Pacific (Melbourne), Europe (Spain), and Europe (Zurich). These vaults are immutable, locked by default, and encrypted with AWS or customer-managed keys. The feature enables cross-account and cross-region backup copying, shared recovery via RAM, and multi-party approval to protect against account compromise.

• Neptune backups can now be stored in logically air-gapped vaults in Melbourne, Spain, and Zurich.
• Vaults are immutable and locked by default to prevent unauthorized modification or deletion.
• Use RAM to share vaults for recovery and multi-party approval to mitigate account compromise risks.
• Deploy via AWS Backup console, CLI, or SDKs to enhance disaster recovery and compliance posture.

Amazon Keyspaces is now available in the Canada West (Calgary) region, enabling Cassandra-compatible applications to run with lower latency while keeping data within the region. This managed, serverless service supports high throughput and storage without operational overhead, allowing teams to use familiar CQL. The expansion helps organizations in Canada meet data residency requirements while scaling applications to thousands of requests per second.

• Cassandra-compatible apps can now deploy in ca-west-1 for reduced latency.
• Serverless model means pay-per-use with virtually unlimited throughput.
• Keeps data within Canada West to satisfy local residency mandates.
• No operational overhead; managed service handles availability and scaling.

Amazon RDS now offers visibility into the initialization progress of storage volumes created from snapshots during point-in-time restores, read replica creation, or Single-AZ to Multi-AZ conversions. This status indicates when blocks downloaded from S3 are fully written and ready for low-latency access. Previously, users faced unpredictable I/O latency during this background initialization process without clear indicators of readiness.

• Monitor initialization status before launching latency-sensitive workloads after restores.
• Eliminate guesswork regarding when provisioned IOPS performance is fully active.
• Applies to point-in-time restores, read replicas, and Single-AZ to Multi-AZ conversions.
• Reduces risk of performance degradation caused by unoptimized block access during init.

Amazon ECS now allows scheduling tasks on partial NVIDIA L4 Tensor Core GPUs using EC2 G6f instances. Users can request specific fractions like 0.125, 0.25, or 0.5 units, which maps to 3 GB to 12 GB of GPU memory. This feature targets small-model AI inference and experimentation workloads that previously required full GPU allocation.

• Set GPU=0.125, 0.25, or 0.5 in ECS task definitions to request fractional resources.
• G6f instances with NVIDIA L4 GPUs are the required underlying hardware.
• Enables right-sizing for small inference models, reducing infrastructure costs.
• Minimum partition size is one-eighth of a GPU, providing 3 GB of memory.

Amazon ElastiCache now supports Graviton4-based M8g, R8g, and C8gn node families for both Valkey and Memcached engines. These new instances deliver up to 47% higher throughput and 43% lower P99 latency compared to Graviton3 equivalents, alongside a 31% improvement in price-performance. The M8g family specifically offers up to 20% more memory per node size, while C8gn nodes provide up to 200 Gbps of network bandwidth.

• Graviton4 nodes boost throughput by 47% and cut P99 latency by 43% over Graviton3.
• M8g nodes provide up to 20% more memory at the same instance size.
• C8gn nodes support 200 Gbps network bandwidth for high-throughput workloads.
• Support covers both Valkey and Memcached engines on Amazon ElastiCache.

AWS Lambda now allows functions running outside a VPC to access network bandwidth that scales proportionally with memory size. Functions configured with 2 GB or more can achieve speeds ranging from 625 Mbps up to 3,000 Mbps at 10 GB. This update addresses prior throughput constraints for latency-sensitive workloads processing large data volumes from external sources.

• Bandwidth scales linearly from 625 Mbps (2 GB) to 3,000 Mbps (10 GB) for VPC-less functions.
• Eliminates previous 625 Mbps bottleneck for large data ingestion workloads.
• Enables faster processing for latency-sensitive tasks transferring terabytes of data.
• Applies specifically to functions not deployed within a VPC configuration.