search

cmu-db / benchbase

Multi-DBMS SQL Benchmarking Framework via JDBC
https://db.cs.cmu.edu/projects/benchbase/
Other
471 stars 185 forks source link
database

readme

BenchBase

BenchBase (Java with Maven)

BenchBase (formerly OLTPBench) is a Multi-DBMS SQL Benchmarking Framework via JDBC.

Table of Contents

Quickstart

To clone and build BenchBase using the postgres profile,

git clone --depth 1 https://github.com/cmu-db/benchbase.git
cd benchbase
./mvnw clean package -P postgres

This produces artifacts in the target folder, which can be extracted,

cd target
tar xvzf benchbase-postgres.tgz
cd benchbase-postgres

Inside this folder, you can run BenchBase. For example, to execute the tpcc benchmark,

java -jar benchbase.jar -b tpcc -c config/postgres/sample_tpcc_config.xml --create=true --load=true --execute=true

A full list of options can be displayed,

java -jar benchbase.jar -h

Description

Benchmarking is incredibly useful, yet endlessly painful. This benchmark suite is the result of a group of PhDs/post-docs/professors getting together and combining their workloads/frameworks/experiences/efforts. We hope this will save other people's time, and will provide an extensible platform, that can be grown in an open-source fashion.

BenchBase is a multi-threaded load generator. The framework is designed to be able to produce variable rate, variable mixture load against any JDBC-enabled relational database. The framework also provides data collection features, e.g., per-transaction-type latency and throughput logs.

The BenchBase framework has the following benchmarks:

This framework is design to allow for easy extension. We provide stub code that a contributor can use to include a new benchmark, leveraging all the system features (logging, controlled speed, controlled mixture, etc.)

Usage Guide

How to Build

Run the following command to build the distribution for a given database specified as the profile name (-P). The following profiles are currently supported: postgres, mysql, mariadb, sqlite, cockroachdb, phoenix, and spanner.

./mvnw clean package -P <profile name>

The following files will be placed in the ./target folder:

How to Run

Once you build and unpack the distribution, you can run benchbase just like any other executable jar. The following examples assume you are running from the root of the expanded .zip or .tgz distribution. If you attempt to run benchbase outside of the distribution structure you may encounter a variety of errors including java.lang.NoClassDefFoundError.

To bring up help contents:

java -jar benchbase.jar -h

To execute the tpcc benchmark:

java -jar benchbase.jar -b tpcc -c config/postgres/sample_tpcc_config.xml --create=true --load=true --execute=true

For composite benchmarks like chbenchmark, which require multiple schemas to be created and loaded, you can provide a comma separated list:

java -jar benchbase.jar -b tpcc,chbenchmark -c config/postgres/sample_chbenchmark_config.xml --create=true --load=true --execute=true

The following options are provided:

usage: benchbase
 -b,--bench <arg>               [required] Benchmark class. Currently
                                supported: [tpcc, tpch, tatp, wikipedia,
                                resourcestresser, twitter, epinions, ycsb,
                                seats, auctionmark, chbenchmark, voter,
                                sibench, noop, smallbank, hyadapt,
                                otmetrics, templated]
 -c,--config <arg>              [required] Workload configuration file
    --clear <arg>               Clear all records in the database for this
                                benchmark
    --create <arg>              Initialize the database for this benchmark
 -d,--directory <arg>           Base directory for the result files,
                                default is current directory
    --dialects-export <arg>     Export benchmark SQL to a dialects file
    --execute <arg>             Execute the benchmark workload
 -h,--help                      Print this help
 -im,--interval-monitor <arg>   Throughput Monitoring Interval in
                                milliseconds
 -jh,--json-histograms <arg>    Export histograms to JSON file
    --load <arg>                Load data using the benchmark's data
                                loader
 -s,--sample <arg>              Sampling window

How to Run with Maven

Instead of first building, packaging and extracting before running benchbase, it is possible to execute benchmarks directly against the source code using Maven. Once you have the project cloned you can run any benchmark from the root project directory using the Maven exec:java goal. For example, the following command executes the tpcc benchmark against postgres:

mvn clean compile exec:java -P postgres -Dexec.args="-b tpcc -c config/postgres/sample_tpcc_config.xml --create=true --load=true --execute=true"

this is equivalent to the steps above but eliminates the need to first package and then extract the distribution.

How to Enable Logging

To enable logging, e.g., for the PostgreSQL JDBC driver, add the following JVM property when starting...

-Djava.util.logging.config.file=src/main/resources/logging.properties

To modify the logging level you can update logging.properties and/or log4j.properties.

How to Release

./mvnw -B release:prepare
./mvnw -B release:perform

How use with Docker

See the docker/benchbase/README.md for further details.

Github Codespaces and VSCode devcontainer support is also available.

How to Add Support for a New Database

Please see the existing MySQL and PostgreSQL code for an example.

Contributing

We welcome all contributions! Please open a pull request. Common contributions may include:

Please see the CONTRIBUTING.md for addition notes.

Known Issues

Please use GitHub's issue tracker for all issues.

Credits

BenchBase is the official modernized version of the original OLTPBench.

The original OLTPBench code was largely written by the authors of the original paper, OLTP-Bench: An Extensible Testbed for Benchmarking Relational Databases, D. E. Difallah, A. Pavlo, C. Curino, and P. Cudré-Mauroux. In VLDB 2014. Please see the citation guide below.

A significant portion of the modernization was contributed by Tim Veil @ Cockroach Labs, including but not limited to:

Citing This Repository

If you use this repository in an academic paper, please cite this repository:

D. E. Difallah, A. Pavlo, C. Curino, and P. Cudré-Mauroux, "OLTP-Bench: An Extensible Testbed for Benchmarking Relational Databases," PVLDB, vol. 7, iss. 4, pp. 277-288, 2013.

The BibTeX is provided below for convenience.

@article{DifallahPCC13,
  author = {Djellel Eddine Difallah and Andrew Pavlo and Carlo Curino and Philippe Cudr{\'e}-Mauroux},
  title = {OLTP-Bench: An Extensible Testbed for Benchmarking Relational Databases},
  journal = {PVLDB},
  volume = {7},
  number = {4},
  year = {2013},
  pages = {277--288},
  url = {http://www.vldb.org/pvldb/vol7/p277-difallah.pdf},
}