Removal of garbage data from a database

A computer program product comprising:

• a computer readable storage medium having computer readable program code embodied therewith and executable by at least one processor, wherein a multi-version concurrency control database stores a database object in a form of a database table as a plurality of different object portions, wherein each object portion includes plural copies of data tuples of the database table and each copy of a data tuple of the database table includes a corresponding change to data of that data tuple of the database table, wherein the plural copies of the data tuples of the database table are stored until the corresponding changes to the data of the data tuples of the database table in the plural copies are reconciled and committed, wherein each copy of a data tuple of the database table is associated with a transaction identifier of a database transaction generating the corresponding change to data for that data tuple of the database table, wherein each transaction identifier corresponds to a time of the corresponding change to data for the associated data tuple of the database table, and wherein the computer readable program code is configured to cause the at least one processor to:
• maintain metadata including garbage collection statistics in the multi-concurrency control database management system indicating visibility of a copy of a data tuple of the database table to a database transaction;
• determine based on the metadata a transaction identifier for an oldest database transaction for which a corresponding copy of the data tuple of the database table is needed and remains visible;
• determine a threshold amount of data to remove for each object portion stored in the multi-version concurrency control database;
• identify a threshold transaction identifier for database transactions for each object portion such that a set of copies of the data tuples in that object portion have associated transaction identifiers older than the threshold transaction identifier based on the time of change of those associated transaction identifiers, and wherein an amount of copies of the data tuples in the set corresponds to the threshold amount of data to remove;
• examine each object portion and determine object portions with the threshold amount of data to remove by comparing the threshold transaction identifiers for database transactions of those object portions to the determined transaction identifier of the oldest database transaction;
• and
• remove data from the database table in response to a sufficient quantity of the determined object portions containing the threshold amount of data to be removed based on the threshold transaction identifiers being at least as old as the determined transaction identifier of the oldest database transaction that is retained in the metadata and is still visible.