MatrixBench, abbreviated as mxbench, is a performance testing and data simulation tool for MatrixDB's streaming data ingestion service. It can rapidly generate random data based on specified parameters such as number of devices, time range, and number of metrics, then insert the data into MatrixDB at high speed via MatrixGate. The tool is located in the MatrixDB installation directory under bin/mxbench.
MatrixBench provides a rich set of parameters to meet various testing requirements. The total number of devices is defined by the tag-range parameter, with generated data evenly distributed across all devices. Timestamps start from ts-begin. For every ts-rps rows inserted, the timestamp advances by one second. The number of metrics per row is controlled by columns-per-row. The metric data type is specified by column-data-type, which can be simple types or complex types such as JSON or Array. For complex types, the individual value type within each column is defined by point-data-type. Below is a detailed list of startup parameters:
| Parameter | Value | Description |
|---|---|---|
| --table-name | tableName | Target table name (required) |
| --database | postgres (default) | Target database name |
| --http-host | localhost (default) | Hostname of MatrixGate |
| --http-port | 8086 (default) | Port number of MatrixGate |
| --parallel | Number of CPU cores (default) | Maximum number of connections to MatrixGate |
| --batch-size | 1000 (default) | Number of data rows per request per connection |
| --tag-range | 50000 (default) | Total number of target devices for data loading |
| --ts-rps | 25000 (default) | Timestamp increment rate; time increases by one second after inserting this many rows |
| --ts-begin | Current time (default) | Start timestamp for generated data |
| --format | text (default) | Data format: text or csv |
| --time-format | unix-second (default) | Timestamp unit: unix-second, unix-ms, unix-nano, or raw |
| --delimiter | | (default) | Column delimiter in each data row |
| --columns-per-row | 100 (default) | Number of columns per data row |
| --column-data-type | simple (default) | Metric column type: supports simple (e.g., float/int), and composite types json, array |
| --points-per-column | 3 (default) | Number of metric values per column when --column-data-type is json or array |
| --point-data-type | float8 (default) | Data type of each metric value within a composite column; supports int, float4, float8 |
| --use-gzip | FALSE (default) | Whether to enable GZIP compression when sending data |
Create a table in the demo database:
create table sensor_data(
"time" timestamp with time zone not null,
tag_id integer,
tag_value double precision)
Distributed By(tag_id);Start MatrixGate:
mxgate config --db-master-port 5432 --db-database demo --interval 150 > ~/mxgate.conf
mxgate start --config ~/mxgate.confRun mxbench:
mxbench --database demo --table-name sensor_data --columns-per-row 1 --column-data-type simple --point-data-type float8After running mxbench, you will see output similar to:
Send to http://127.0.0.1:8086
Parallel level is 64
Point format is simple
Tag range is 50000
TIMESTAMP step is 40000 nano second
16036-03-16 16:55:04: sent OK: 4282000, ERR 0, Lat 147.67ms last10s 147.67ms, Row/s: 428200, last10s 428200
16036-03-16 16:55:14: sent OK: 8506000, ERR 0, Lat 148.12ms last10s 148.58ms, Row/s: 425300, last10s 422400
16036-03-16 16:55:24: sent OK: 12794000, ERR 0, Lat 148.26ms last10s 148.52ms, Row/s: 426466, last10s 428800OK: Total number of rows successfully sent and inserted since mxbench started.ERR: Total number of failed rows.Lat: Average latency since startup.last10s after Lat: Average latency over the last 10 seconds.Row/s: Average number of rows sent per second since startup.last10s after Row/s: Average send rate over the last 10 seconds.Once started, mxbench continues load testing until the user presses Ctrl+C to terminate the process.