Key-Value load

A simple load type using a YDB database as a Key-Value storage.

Types of load

This load test runs several types of load:

• upsert: Using the UPSERT operation, inserts rows that are tuples (key1, key2, ... keyK, value1, value2, ... valueN) into the table created previously with the init command, the K and N numbers are specified in the settings.
• insert: The function is the same as the upsert load, only the INSERT operation is used for insertion.
• select: Reads data using the SELECT * WHERE key = $key operation. A query always affects all table columns, but isn't always a point query, and the number of primary key variations can be controlled using parameters.
• read-rows: Reads data using the ReadRows operation, which performs faster key reading than select operation. A query always affects all table columns, but isn't always a point query, and the number of primary key variations can be controlled using parameters.
• mixed: Simultaneously writes and reads data, additionally checking that all written data is successfully read.

Load test initialization

To get started, you must create tables. When creating them, you can specify how many rows to insert during initialization:

ydb workload kv init [init options...]

• init options: Initialization options.

View a description of the command to initialize the table:

ydb workload kv init --help

Available parameters

The following command is used to create a table:

CREATE TABLE `kv_test`(
    c0 Uint64,
    c1 Uint64,
    ...
    cI Uint64,
    cI+1 String,
    ...
    cN String,
    PRIMARY KEY(c0, c1, ... cK)) WITH (
        AUTO_PARTITIONING_BY_LOAD = ENABLED,
        AUTO_PARTITIONING_MIN_PARTITIONS_COUNT = partsNum,
        UNIFORM_PARTITIONS = partsNum,
        AUTO_PARTITIONING_PARTITION_SIZE_MB = partSize,
        AUTO_PARTITIONING_MAX_PARTITIONS_COUNT = 1000
    )
)

Examples of load initialization

Example of a command to create a table with 1000 rows:

ydb workload kv init --init-upserts 1000

Deleting a table

When the work is complete, you can delete the table:

ydb workload kv clean

The following YQL command is executed:

DROP TABLE `kv_test`

Examples of using clean

ydb workload kv clean

Running a load test

To run the load, execute the command:

ydb workload kv run [workload type...] [global workload options...] [specific workload options...]

During this test, workload statistics for each time window are displayed on the screen.

• workload type: The types of workload.
• global workload options: The global options for all types of load.
• specific workload options: Options of a specific load type.

See the description of the command to run the data load:

ydb workload kv run --help

Global parameters for all types of load

Upsert load

This load type inserts tuples (key, value1, value2, ..., valueN)

To run this type of load, execute the command:

ydb workload kv run upsert [global workload options...] [specific workload options...]

• global workload options: The global options for all types of load.
• specific workload options: Options of a specific load type.

For example, for the parameters --rows 2 --cols 3 --int-cols 2, the YQL query will look like this:

DECLARE $c0_0 AS Uint64;
DECLARE $c0_1 AS Uint64;
DECLARE $c0_2 AS String;
DECLARE $c1_0 AS Uint64;
DECLARE $c1_1 AS Uint64;
DECLARE $c1_2 AS String;
UPSERT INTO `kv_test` (c0, c1, c2) VALUES ($c0_0, $c0_1, $c0_2), ($c1_0, $c1_1, $c1_2)

Parameters for upsert

Insert load

This load type inserts tuples (key, value1, value2, ..., valueN)

To run this type of load, execute the command:

ydb workload kv run insert [global workload options...] [specific workload options...]

• global workload options: The global options for all types of load.
• specific workload options: Options of a specific load type.

For example, for the parameters --rows 2 --cols 3 --int-cols 2, the YQL query will look like this:

DECLARE $c0_0 AS Uint64;
DECLARE $c0_1 AS Uint64;
DECLARE $c0_2 AS String;
DECLARE $c1_0 AS Uint64;
DECLARE $c1_1 AS Uint64;
DECLARE $c1_2 AS String;
INSERT INTO `kv_test` (c0, c1, c2) VALUES ($c0_0, $c0_1, $c0_2), ($c1_0, $c1_1, $c1_2)

Parameters for insert

Select load

This type of load creates SELECT queries that return rows based on an exact match of the primary key.

To run this type of load, execute the command:

ydb workload kv run select [global workload options...]

• global workload options: The global options for all types of load.

For example, for the parameters --rows 2 --cols 3 --int-cols 2, the YQL query will look like this:

DECLARE $r0_0 AS Uint64;
DECLARE $r0_1 AS Uint64;
DECLARE $r1_0 AS Uint64;
DECLARE $r1_1 AS Uint64;
SELECT c0, c1, c2 FROM `kv_test` WHERE c0 = $r0_0 AND c1 = $r0_1 OR c0 = $r1_0 AND c1 = $r1_1

Read-rows load

This type of load creates ReadRows queries that return rows based on an exact match of the primary key.

To run this type of load, execute the command:

ydb workload kv run read-rows [global workload options...]

• global workload options: The global options for all types of load.

Mixed load

This type of load simultaneously writes and reads tuples (key, value1, value2, ..., valueN), additionally checking that all written data is successfully read.

To run this type of load, execute the command:

ydb workload kv run mixed [global workload options...] [specific workload options...]

• global workload options: The global options for all types of load.
• specific workload options - Options of a specific load type.

Parameters for mixed