Abstract: Disclosed herein are systems and methods for indexing multi-versioned data. In one aspect, a method includes: generating a first index for a first snapshot of a collection of data objects in a volume captured at a first time, wherein each data object is identified by a respective key; generating a second index for a second snapshot of the volume captured at a second time, wherein the second index includes a subset of sorted runs from the first index and one or more additional sorted runs; combining the first and the second index to form a multi-versioned log-structured object index that has sorted runs from both indexes; and wherein the combining further includes generating a global directory of the multi-versioned log-structured object index that identifies each snapshot and corresponding sorted runs.