Skip to content
This repository has been archived by the owner on Jun 12, 2020. It is now read-only.

TPC H load and check test

prohaska edited this page Sep 19, 2014 · 3 revisions

TPC-H load and check test

The run.tpch.bash script loads a TPC-H database of a given scale factor, checks the table for correctness, dumps the table, and compares the dump to a dump of the same database that was stored in InnoDB. The test also builds a clustering index on the 'lineitem' table.

The test expects that a mysql build is installed in '/usr/local' and that mysqld is running on unix socket '/tmp/mysql.sock'. There are command line arguments that control

  • whether or not the loader is used to load the data (tokudb_prelock_empty),
  • whether or not the loader compresses intermediate files (tokudb_load_save_space),
  • the compression algorithm used for the table (tokudb_row_format),
  • the amount of memory used for the loader (tokudb_loader_memory_size).

The test results are stored in Tokutek's subversion source control system.

See the tpch.readme for more details.

Expected results: all tests pass.

Future work

  • There are a set of data warehouse like queries that can be run on the TPC-H database. These queries are not currently run because
    • the MySQL query optimizer does not always generate a good plan for the query,
    • MySQL does no query parallelization, so the execution time is huge.
  • Run the queries on small scale TPC-H databases and compare the results to the expected result set.
  • The table load times can be compared with the load times of previous test runs.
Clone this wiki locally