Benchmark Tests

Running the Benchmarks

Syntax for running run.py:

$ ./run.py [-l -v <log_level> -s <run_speed> -u <user> -p <password> -i <instance>] [Benchmark1 ... BenchmarkN]

Specifying a specific benchmark or set of benchmarks runs only those, while not specifying any runs all benchmarks.

-l lists the benchmarks that will be run
-v <run_speed> can either be slow, medium or fast
-s <log_level> is a number representing the verbosity of the debugging output: 10 is debug, 20 is info, 30 is warning, etc.
-u <user> user to use when connecting with accumulo. If not set you will be prompted to input it.
-p <password> password to use when connecting with accumulo. If not set you will be prompted to input it.
-z <zookeepers> comma delimited lit of zookeeper host:port pairs to use when connecting with accumulo. If not set you will be prompted to input it.
-i <instance> instance to use when connecting with accumulo. If not set you will be prompted to input it.

The Benchmarks

Values in a 3-tuple are the slow,medium,fast speeds at which you can run the benchmarks.

• CloudStone1: Test the speed at which we can check that accumulo is up and we can reach all the tservers. Lower is better.
• CloudStone2: Ingest 10000,100000,1000000 rows of values 50 bytes on every tserver. Higher is better.
• CloudStone3: Ingest 1000,5000,10000 rows of values 1024,8192,65535 bytes on every tserver. Higher is better.
• CloudStone4 (TeraSort): Ingests 10000,10000000,10000000000 rows. Lower score is better.
• CloudStone5: Creates 100,500,1000 tables named TestTableX and then deletes them. Lower is better.
• CloudStone6: Creates a table with 400, 800, 1000 splits. Lower is better.

Terasort

The 4th Benchmark is Terasort. Run the benchmarks with speed ‘slow’ to do a full terasort.

Misc

These benchmarks create tables in accumulo named ‘test_ingest’ and ‘CloudIngestTest’. These tables are deleted at the end of the benchmarks. The benchmarks will also alter user auths while it runs. It is recommended that a benchmark user is created.