S3 Tables now come with tagging, Amazon announced last week, making it possible to run cost collocation and attribute-based access control on S3's newest bucket.
Tagging is one of several recent new features for Tables, alongside general quality-of-life updates to S3, as the open source competition builds around what appears to be a strategically important product for the hyperscaler.
AWS launched Tables in December 2024 as the Simple Storage Service's third bucket type (after directory buckets and general purpose buckets).
"You can think of a table bucket as an analytics warehouse that can store Iceberg tables with various schemas." - AWS
AWS CEO Matt Garman said Tables was created in part to make S3 better for large analytics and AI workloads.
"You can think of a table bucket as an analytics warehouse that can store Iceberg tables with various schemas," said AWS, promising much better query throughput.
Tables is built to handle Apache Iceberg, the open-source table format for big data, and is being help up by Amazon as proof of innovation around data lakes. Iceberg has seen a surge in adoption for real-time AI applications.
Cost and lock-in
But Tables received a mixed reception, with warnings about unpredictable costs and the potential for vendor lock-in.
"Amazon S3 Tables aren’t S3 or a data lake, they are in fact a proprietary table API for reading and writing tables; not reading and writing files," wrote data engineer Daniel Beach.
In July, AWS announced that it was dropping the price of table compaction by between 50% (for per-object) and 90% (for binpack compaction).
Cost has been a recurring theme in S3 Tables criticism, as well as the top use case listed for tags in Tables, as it has been in other buckets.
S3 cost allocation tagging – at a bucket level – was added to the API in 2012, and object tagging followed in 2016.
Once Tables tags are set they can be activated in bulk via API, and then tracked in both the AWS cost allocation report and the cost and usage report.
Tags can also be used for attribute-based access control (ABAC), allowing access to S3 resources based on their tags.
Bread-and-butter updates
In late October, S3 gained conditional write for copy operations, and started to log automatic S3 Tables maintenance, a big part of the pitch for Tables, in AWS CloudTrail.
Since late September, S3 Tables can be previewed in the S3 console without a SQL query, and in mid September, AWS rolled out a deletion equivalent of conditional-write copy: conditional deletes in S3 general purpose buckets.
Also in September, AWS for the first time enabled S3 batch operations to specify not just objects, but entire buckets or storage classes, or to filter by attributes such as creation date.
The link has been copied!
