BenchBase
BenchBase (formerly OLTPBench) is a Multi-DBMS SQL Benchmarking Framework via JDBC.
Table of Contents
Quickstart
To clone and build BenchBase,
git clone --depth 1 https://github.com/cmu-db/benchbase.git
cd benchbase
./mvnw clean package
This produces artifacts in the target folder, which can be extracted,
cd target
tar xvzf benchbase-x-y-z.tgz # Change x-y-z appropriately.
cd benchbase-x-y-z # Change x-y-z appropriately.
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:
(TODO(WAN): Get permission to migrate/copy Tim's wiki for benchmark descriptions)
- AuctionMark
- CH-benCHmark
- Epinions.com
- hyadapt -- pending configuration files
- NoOp
- Resource Stresser
- SEATS
- SIBench
- SmallBank
- TATP
- TPC-C
- TPC-H
- TPC-DS -- pending configuration files
- Voter
- Wikipedia
- YCSB
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:
./mvnw clean package
The following files will be placed in the ./target folder:
benchbase-x.y.z.tgzbenchbase-x.y.z.zip
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: oltpbenchmark
-b,--bench <arg> [required] Benchmark class. Currently
supported: [tpcc, tpch, tatp, wikipedia,
resourcestresser, twitter, epinions, ycsb,
seats, auctionmark, chbenchmark, voter,
sibench, noop, smallbank, hyadapt]
-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
--load <arg> Load data using the benchmark's data
loader
-s,--sample <arg> Sampling window
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 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:
- Adding support for a new DBMS.
- Adding more tests of existing benchmarks.
- Fixing any bugs or known issues.
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:
- Built with and for Java 11.
- Migration from Ant to Maven.
- Reorganized project to fit Maven structure.
- Removed static
libdirectory and dependencies. - Updated required dependencies and removed unused or unwanted dependencies.
- Moved all non
.javafiles to standard Mavenresourcesdirectory. - Shipped with Maven Wrapper.
- Improved packaging and versioning.
- Moved to Calendar Versioning (https://calver.org/).
- Project is now distributed as a
.tgzor.zipwith an executable.jar. - All code updated to read
resourcesfrom inside.jarinstead of directory.
- Moved from direct dependence on Log4J to SLF4J.
- Reorganized and renamed many files for clarity and consistency.
- Applied countless fixes based on "Static Analysis".
- JDK migrations (boxing, un-boxing, etc.).
- Implemented
try-with-resourcesfor alljava.lang.AutoCloseableinstances. - Removed calls to
printStackTrace()orSystem.out.printlnin favor of proper logging.
- Reformatted code and cleaned up imports.
- Removed all calls to
assert. - Removed various forms of dead code and stale configurations.
- Removed calls to
commit()duringLoaderoperations. - Refactored
WorkerandLoaderusage ofConnectionobjects and cleaned up transaction handling. - Introduced
HikariCPas connection pool andDataSourceinstead of building connections fromDriverManageras needed (defaultpoolsizeis 12). - Introduced Dependabot to keep Maven dependencies up to date.
- Simplified output flags by removing most of them, generally leaving the reporting functionality enabled by default.
- Provided an alternate
Catalogthat can be populated directly from the configured Benchmark database. The old catalog was proxied throughHSQLDB-- this remains an option for DBMSes that may have incomplete catalog support.
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},
}
1.1k Dec 28, 2022
3 Mar 28, 2022
17.7k Jan 1, 2023
94 Apr 20, 2022
1.1k Jan 1, 2023
5 Apr 7, 2022
547 Dec 25, 2022
1 Jan 27, 2022
6 Oct 26, 2022
6 Oct 14, 2022
19 Nov 7, 2022
990 Dec 27, 2022
21 Dec 9, 2022
2 Mar 8, 2022
6.9k Dec 31, 2022
3.1k Jan 5, 2023
14.3k Dec 30, 2022
5.3k Jan 4, 2023
3.6k Jan 2, 2023