Scriptable Neo4j benchmarks. Neobench helps you create and run artificial load to tune your Neo4j deployments.

• Benchmark throughput and latency
• Output in easy-to-process CSV
• Configurable concurrency
• Allows mixed workloads
• Built-in TPC-B and LDBC SNB benchmarking modes
• Custom workloads using built-in scripting language

You can download binaries in the "assets" section in the latest release here.

Easier to use if you don't want to run against a database running on the host machine.

docker run jjdh/neobench -h

Run neobench -h for a list of available options.

# Run the built-in "TPC-B-like" benchmark in throughput testing mode.
# Before running the benchmark, run the built-in dataset populator for TPC-B (--init)
$ neobench --address neo4j://localhost:7687 --password secret \
    --builtin tpcb-like --init

See the Docs for more details.

Build and run integration tests with make

Release with bin/release

• Functions should be changed to align with cypher functions as much as possible (eg. greatest, least etc)
• Syntax around parameters and when to use $ and when not to is confusing in list comprehensions, maybe default to never using $ except in queries? Eg :set a [ i in range(1,10) | $i ]; why no $ the first time i is used, but then $ when it's used the second time?
• Load generation system should probably be modified to have a single goroutine generate load that goes on a queue for workers to execute, rather than each worker generating its own load, because that causes weird blocking issues
• Support dropping transactions / marking them failed when database can't keep up with a set rate
• Warn or crash if user specifies --rate without also specifying --latency

This project has no current maintainer.

Apache 2