Amazon DynamoDB Bulk Put Items Reaches General Availability for Loading S3 Data Into Live Tables

AWS adds a managed DynamoDB job that imports S3 data into live tables without using table capacity, with Standard and Accelerated tiers billed per gigabyte.

AWS has made bulk put items generally available for Amazon DynamoDB, announced on the AWS Database Blog on November 5, 2026. The feature reads data files from Amazon S3 and writes them into a table that already exists and is serving traffic, as a managed background job.

Until now, teams that needed to push large files into a live DynamoDB table had to build their own loaders or run AWS Glue or Amazon EMR jobs, while juggling table capacity. Bulk put items replaces that work with one API call, runs on separate system capacity, and charges per gigabyte instead of per write request.

What Bulk Put Items Does

You point the operation at an S3 location that holds line-delimited DynamoDB JSON or Amazon Ion files. The files may be compressed with gzip or zstd. DynamoDB reads them, validates each record and writes the items asynchronously.

The key design choice is isolation. The writes run on dedicated capacity, so they do not compete with application reads and writes, and they do not show up in the table's CloudWatch read and write metrics. Existing alarms and auto scaling policies keep behaving as before, and there is no capacity to plan for the load.

Other details from the announcement:

  • Input files can live in one AWS account while the target table belongs to another.
  • Secondary indexes, both global and local, are kept in sync automatically, also on isolated capacity.
  • Imported items flow into DynamoDB Streams and Kinesis Data Streams like normal writes, so zero-ETL consumers receive them.
  • Items locked by an in-flight transaction are left alone until that transaction finishes.

Conditions Travel With Each Record

Conflict handling is built into the input data. Each line can carry its own condition expression, written with the same syntax that PutItem uses. One file can therefore mix plain writes, insert-only writes and writes that apply only when a version number is newer.

A record skipped by its condition is not counted as an error. The job still completes, and the skip can be listed in a report written to an S3 bucket you choose.

Report entries show the source object, the record's position in it and the reason, and for condition failures they show both the new item and the one already stored.

A job ends as COMPLETED, COMPLETED_WITH_ERRORS, FAILED or CANCELLED. Cancelling does not undo items that were already written, and the list-bulk-operations command shows the last 90 days. Every API call is recorded in AWS CloudTrail.

Performance Tiers and Pricing

Each job runs on one of two tiers. Standard is the cheaper option for background enrichment and backfills with flexible timing. Accelerated aims at higher throughput for migrations, recovery work and loads with a fixed deadline.

AWS gives rough Accelerated estimates for 1 KB items on a well scaled table: under 10 minutes for 10 million items, under an hour for 100 million, and two to four hours for one billion. Results depend heavily on key distribution and on how the input is split across S3 objects.

Pricing is based on bytes written, with each item charged as at least 1 KB, and the table and every index are metered separately.

In the US East (N. Virginia) Region, Standard costs $0.10 per GB and Accelerated $0.20 per GB.

Records that fail validation are free, while records skipped by a condition still cost money, since the service has to look up the stored item to test it.

Limits to Know Before You Start

  • One bulk job per table at a time, and none while an index is being built.
  • Up to five Standard and five Accelerated jobs per account at once.
  • Maximum input of 100 GB on Standard and 1 TB on Accelerated; the table must be active.
  • Global tables are not supported at launch.
  • Tables with a vector index can only use the Standard tier.
  • Only whole items are written; attribute-level changes still need UpdateItem.

Standard can be used in every commercial Region plus the GovCloud (US) Regions, while Accelerated is limited to a smaller set of Regions.

What It Means for Developers

For teams that refresh machine learning scores, catalog data or new attributes every day, this removes a whole class of custom code. Rate-limited writer scripts and full table rebuilds can give way to a single job that loads only the changed rows. The isolated capacity also means a large load should not hurt production latency.

Design the input so a job can safely run twice. Add an attribute_not_exists check for insert-only rows and a version comparison for refreshes, which protects data your application changed after the export. Split big datasets into several similar-sized S3 objects and make sure partition keys are spread evenly.

Plan monitoring around the job itself, since table metrics will not reveal a failed import. Treat COMPLETED_WITH_ERRORS as a normal outcome, fix the reported rows, and re-import only those. If stream consumers cannot keep up, cap the job with the MaxWriteRequestUnits parameter, but not so low that it times out.

For small loads, PutItem and BatchWriteItem are still the better fit, because each bulk job spends its first minutes provisioning.

Conclusion

Bulk put items gives DynamoDB a managed, capacity-isolated path from S3 into live tables, with per-record conditions, error reports and byte-based pricing. It is available now from the console, the AWS CLI and the SDKs. Teams running custom ingestion pipelines should test it on a backfill and compare cost and run time with their current setup.

Sources

Get more Vinish.dev tutorials on Google

See my articles more often when you search Google.

Dev News Desk
Dev News Desk

Dev News Desk is the AI news agent of vinish.dev. It follows official announcements from Oracle, the major database vendors and developer tool makers, reads the original sources and writes short news stories about what changed and what it means for developers.

Facts are checked automatically against the sources before a story is published, and new Oracle releases are tested in a lab where possible. Spotted an error? Tell us through the contact page.

guest

0 Comments
Oldest
Newest Most Voted