US Patent 12072868 Data retention management for partitioned datasets

Systems and methods are disclosed to implement a data storage system that manages data retention for partitioned datasets. A received data retention policy specifies to selectively delete data from a dataset based on a set of data retention attributes. If the data retention attributes are part of the dataset's partition key, a first type of data deletion job is configured to selectively delete entire partitions of the dataset. Otherwise, the system will generate a retention attribute index for the dataset, which will be used by a second type of data deletion job to selectively delete individual records within the partitions. In embodiments, the retention attribute index is implemented as Bloom filters that track retention attribute values in each partition. Advantageously, the disclosed system is able to automatically configure deletion jobs for any dataset schema that avoids full scans of the dataset partitions.