membench

Membench is the tool to benchmark in-mem databases like DiceDB, Redis, etc for various operations.

To run membench, build the membench from source

$ make build

Running Membench

$ ./membench benchmark --database dicedb \
    --host localhost \
    --port 7379 \
    --num-requests 100000 \
    --num-clients 4

You can always get help by running

$ ./membench benchmark --help

Telemetry Sink

Membench supports multiple telemetry sinks like

• mem - which accumulates the stats in an in-mem
• prometheus - which emits the stats to a prometheus instance

You can configure this using the flag --telemetry-sink.

Memory - Telemetry Sink

Accumulates all the metrics in in-memory histograms and outputs the report at the end of the benchmark.

op,latency_ns_avg,latency_ns_p50,latency_ns_p90,latency_ns_p95,latency_ns_p99
GET,82647,71679,103935,130047,264191
SET,89865,73215,107007,134143,290815
op,error_count
GET,0
SET,0

Prometheus - Telemetry Sink

To make it simpler, there is a docker-compose.yml file that starts prometheus and grafana. You can run the following command to start the telemetry stack

$ docker compose up

If you are using prometheus, then make sure you are updating the file prometheus.yml and setting the correct IP address.

You can get the IP address of the machine using the following command

$ ip route | sed -n '2p' | awk '{print $NF}'

You can also use grafana.json file and load it to visualize the membench as it runs with all key vitals.