Which feature is used to roll back a corrupted HDFS instance to a previously known good point in time?
replication
high availability
snapshot
partitioning