sysbench

Multi-threaded system benchmark tool

TLDR

CPU benchmark

$ sysbench cpu run

CPU benchmark with threads

$ sysbench cpu --threads=[4] run

Memory benchmark

$ sysbench memory run

File I/O prepare

$ sysbench fileio --file-total-size=[10G] prepare

File I/O benchmark

$ sysbench fileio --file-total-size=[10G] --file-test-mode=[rndrw] run

MySQL benchmark

$ sysbench oltp_read_write --mysql-host=[localhost] --mysql-user=[user] --mysql-password=[pass] --mysql-db=[test] run

Cleanup test files

$ sysbench fileio --file-total-size=[10G] cleanup

CPU benchmark with custom prime limit

$ sysbench cpu --cpu-max-prime=[20000] --threads=[4] run

SYNOPSIS

sysbench test [options] [prepare|run|cleanup]

sysbench is a multi-threaded benchmarking tool for evaluating system performance across CPU, memory, file I/O, and database workloads. It provides consistent, reproducible measurements that allow meaningful comparisons between different hardware configurations, kernel versions, or tuning parameters.The tool includes built-in test types for common benchmarks. CPU tests stress processors using prime number calculations to measure computational throughput. Memory tests measure bandwidth and latency across sequential and random access patterns. File I/O tests support multiple modes including sequential writes, sequential reads, and random read/write mixes, requiring a preparation step to create test files before running and a cleanup step afterward.Database benchmarks simulate OLTP (Online Transaction Processing) workloads against MySQL or PostgreSQL, executing realistic mixes of select, insert, update, and delete queries. All test types support multi-threaded execution to measure how performance scales with parallelism, revealing bottlenecks and contention points. Custom Lua scripts can define additional workloads beyond the built-in tests.

PARAMETERS

cpu

CPU test.

memory

Memory test.

fileio

File I/O test.

oltp_read_write

Database OLTP test.

oltp_read_only

Read-only database test.

threads

Thread scheduler performance test.

mutex

Mutex contention performance test.

--threads N

Number of threads (default: 1).

--time SECONDS

Test duration (default: 10).

--events N

Maximum number of events (default: 0, unlimited).

--file-total-size SIZE

Total file size.

--file-test-mode MODE

I/O mode (seqwr, seqrd, rndrw, etc.).

--mysql-host HOST

MySQL host.

--mysql-user USER

MySQL user.

--mysql-password PASS

MySQL password.

--mysql-db DB

MySQL database name.

--cpu-max-prime N

Upper limit for prime number calculation in CPU test (default: 10000).

--memory-block-size SIZE

Size of memory block for memory test (default: 1K).

--memory-total-size SIZE

Total size of data to transfer in memory test (default: 100G).

--memory-oper OP

Memory operation type: read or write (default: write).

--memory-access-mode MODE

Memory access mode: seq or rnd (default: seq).

--file-num N

Number of files to create (default: 128).

--file-extra-flags FLAGS

Additional flags for opening files: sync, dsync, direct.

--percentile N

Percentile to calculate in latency statistics (default: 95).

--report-interval SECONDS

Periodically report intermediate statistics (0 = disabled).

prepare

Prepare test data (required for fileio and database tests).

run

Run the benchmark test.

cleanup

Clean up test data created during prepare.

CAVEATS

Results depend on test parameters. Synthetic benchmarks don't reflect all workloads. File tests need disk space. Database tests need setup.

HISTORY

sysbench was created by Alexey Kopytov at MySQL AB around 2004. It became a standard tool for MySQL benchmarking and general system performance testing.