AWS amazondynamodb documentation change
Summary
Added documentation about segment assignment mechanics during parallel scans based on partition key hashing
Security assessment
The changes clarify technical scan behavior and performance considerations. No references to security controls, vulnerabilities, or access management. The text focuses on data distribution patterns during scan operations.
Diff
diff --git a/amazondynamodb/latest/developerguide/Scan.md b/amazondynamodb/latest/developerguide/Scan.md index 69844c046..b3632caa0 100644 --- a//amazondynamodb/latest/developerguide/Scan.md +++ b//amazondynamodb/latest/developerguide/Scan.md @@ -216,0 +217,4 @@ In this diagram, the application spawns three threads and assigns each thread a +DynamoDB assigns items to _segments_ by applying a hash function to each item's partition key. For a given `TotalSegments` value, all items with the same partition key are always assigned to the same `Segment`. This means that in a table where _Item 1_ , _Item 2_ , and _Item 3_ all share `pk="account#123"` (but have different sort keys), these items will be processed by the same worker, regardless of the sort key values or the size of the _item collection_. + +Because _segment_ assignment is based solely on the partition key hash, segments can be unevenly distributed. Some segments might contain no items, while others might contain many partition keys with large item collections. As a result, increasing the total number of segments does not guarantee faster scan performance, particularly when partition keys are not uniformly distributed across the keyspace. +