In the latest AWS announcements, the Bedrock AgentCore runtime received a V2 upgrade, S3 Tables now support managed migration of Hive tables to Iceberg, DynamoDB added bulk‑put loading from S3, and a new Web Answer tool was added to AgentCore gateways. Each of these updates alters how AI, cloud, and DevOps teams provision resources, move data, and design low‑latency services.
AgentCore Runtime V2: Memory‑Efficient, Faster Cold Starts
The new runtime is enabled by setting platformVersion to V2 when creating or updating an AgentCore runtime. Memory is allocated on demand: sessions start small, page in additional memory as needed, and release it when idle, which can reduce GB‑hour consumption. Cold‑start latency is improved by snapshotting a prepared environment and restoring it for each new instance; internal tests showed a P75 cold start around 2 seconds for images from 200 MB up to 2 GB, compared with the original runtime’s 5.4 seconds to nearly 30 seconds as image size grew. The update process takes several minutes while the snapshot is built, and the container must respond to /ping within 120 seconds to be considered healthy. V2 is currently available in ten regions, with baseline pricing announced separately.
Managed Hive‑to‑Iceberg Migration with S3 Tables
S3 Tables now let you move Parquet tables registered in the Glue Data Catalog from Apache Hive to Apache Iceberg without building a custom pipeline. The service validates the source, generates a compatibility report, and performs incremental syncs that copy only changed objects, allowing production workloads to continue writing until cut‑over. After migration, the table can be queried via Athena, SageMaker Unified Studio, or any client that supports the Iceberg REST API. Costs are limited to the data copied; there is no charge for the conversion itself, and existing S3 storage class tiers are preserved. The feature launches in the same account and region, initially in ten regions.
Bulk‑Put Items for DynamoDB: Large‑Scale S3‑Based Loads
DynamoDB now accepts line‑delimited DynamoDB JSON or Amazon Ion files stored in S3 and loads them into existing tables as an asynchronous job. Each record may include its own condition expression, enabling mixed write patterns (plain inserts, version‑guarded updates) within a single file. The operation uses dedicated system capacity, which is billed at a lower rate than per‑request writes. An Accelerated tier is available for time‑critical jobs; the accompanying blog post cites sub‑hour completion for 100 million 1 KB items. Practitioners should note that cancelling a job does not roll back items already written.
Web Answer on AgentCore Gateways
The new Web Answer tool provides a single callable function that returns a synthesized, cited answer to a natural‑language query. It performs its own parallel searches, aggregates evidence, and respects optional domain include/deny lists and publish‑date filters. The tool is attached as a target on an AgentCore gateway and invoked over the MCP protocol. It is currently available in US East (N. Virginia) and is positioned as a higher‑level alternative to building a custom search‑retrieve‑synthesize pipeline.
Related CloudNinjas coverage: AWS.
What This Means For Practitioners
Engineers should evaluate existing AgentCore deployments and consider opting into V2 by updating the platformVersion field, ensuring containers meet the 120‑second health check requirement. For data lake migrations, S3 Tables offers a low‑effort path to Iceberg that can be incorporated into cut‑over plans without halting production writes. DynamoDB bulk‑put should be used for batch ingestion workloads, with attention to the non‑rollback behavior on cancellation. Web Answer can simplify the addition of citation‑rich Q&A capabilities, but teams must assess its regional availability and filter configuration limits. Finally, keep an eye on the expanding regional support for these features and any forthcoming pricing details that could affect cost modeling.