These benchmarks exercise storage apis in a real-life usage way (avoiding microbenchmarks).
This models an offline load of a Google doc. See this document for a breakdown of the database and the transactions, along with the traces used to extract this information.
This benchmark models the creation and reading of a large number of blobs. The blobs are created first, then read one at a time.
This benchmark models the creation and reading of a large number of blobs. The blobs are created first, then read all at once.
This benchmark models the creation and reading of a large number of blobs. Each blob is read immediately after creation.