Uber Technologies, Inc.
Checksum tree generation for improved data accuracy verification

Abstract:

A data management system verifies the accuracy of data retrieved from a primary data store using a checksum tree stored by a secondary data store. A checksum tree is a tree graph that represents a hierarchy of checksums. Leaf nodes of the checksum tree can store checksums for data blocks stored by the primary data store and secondary data store, and parent nodes can represent checksums of their respective child nodes. The data management system can compare reference subtrees within the checksum tree to comparison subtrees that are generated from data retrieved from the primary data store to determine whether the retrieved data is accurate. The data management system can also use the checksum tree to identify which, if any, of the retrieved data blocks are inaccurate.