Kunpeng AutoTuner Functions

Based on the 10 supported big data and database applications, the tool leverages built-in AI algorithms to obtain parameter settings that deliver superior performance, and then applies these settings in actual applications to improve performance.

When Kunpeng 920 series processors are used, Kunpeng register parameters are automatically tuned. The OS kernel version must be 4.19.90-2012.4.0.0053.oe1.aarch64 or 5.10.0-182.0.0.95.oe2203sp3.aarch64.

Figure 1 Kunpeng AutoTuner process

**Table 1** Kunpeng AutoTuner application scenarios
Scenario		Application Startup Method	Pressure Test Method	Description
Database	MySQL	Using the configuration file specified by mysqld	Using Sysbench	You can modify the simple tuning template generated by the tool and use it for tuning. For details about each scenario, see 3.
	openGauss	Using the database folder specified by gs_ctl	Using BenchmarkSQL
	Vastbase	Using the database folder specified by vb_ctl	Using BenchmarkSQL
	RocksDB	Using db_bench	Using rocksdb_dbbench
	PostgreSQL	Using the data directory specified by pg_ctl	Using BenchmarkSQL
	Redis	Using the data directory specified by redis-server	Use the built-in redis-benchmark tool
Big data	Hive	Using the Hive executable file	Using tpcds
	Spark	Automatically loading the configuration file on yarn mode: Use yarn-session.sh to submit a task for tuning. standalone mode: Use start-all.sh to start.	Using tpcds
	Flink	Automatically loading the configuration file on yarn mode: Use yarn-session.sh to submit a task for tuning. standalone mode: Use start-cluster.sh to start.	Using HiBench
	Kafka	Using the configuration file specified by kafka-server-start.sh	Using the built-in kafka-producer-perf-test.sh file
Custom	Custom	-	-	You can modify the custom tuning template generated by the tool and use it for tuning.
	Redis	Using the data directory specified by redis-server	Using memtier_benchmark	The tool provides an additional custom tuning template. You can modify this template for tuning. If the tool is installed using an RPM package, the template file is stored in the /usr/local/devkit/kat/template_extend folder. If the tool is installed using a compressed package, the template file is stored in the ./kat/template_extend folder in the directory where the CLI tool is decompressed.
	Spark	Automatically loading the configuration file on yarn mode: Use yarn-session.sh to submit a task for tuning. standalone mode: Use start-all.sh to start.	Using HiBench

Prerequisites

You have installed the Kunpeng AutoTuner. See Installing the Tool.
If you have installed the tool using a compressed package, decompress the package and switch to the tool directory. Then run the command in ./ mode, for example, ./devkit kat -h. If you have installed the tool using an RPM package, run devkit kat -h. This section uses an RPM package as an example.

Command Function

Automatically tunes application parameters to improve application performance metrics in different scenarios.

Syntax

      
            devkit kat [-h | --help] TASK [ARGS]

Complete the following settings before starting a task:

Task parameters: parameter name, value range, and setting mode.
Application scenario: application name, benchmark (if any), and performance metrics. Ensure that the application and performance test tool work properly.

Adjust the task parameters and application scenario based on the actual service scenario and the parameters to be tuned.

Example

Run the following command to view the information about the functions supported by the AutoTuner:

      
           devkit kat -h

Command output:

      
           Usage: devkit kat [-h | --help] TASK [ARGS]

 The most commonly used devkit kat sub tasks are:
   help       Get help information
   train      Run the auto tuner train task
   template   Run the auto tuner template task
   use        Run the auto tuner use task
   man        Run the auto tuner man task

 See 'devkit kat TASK --help' for more information on a specific task.

**Table 2** Function description
Function	Description
help	Displays help information.
train	Enables automatic tuning.
template	Generates a template file.
use	Uses the tuning result.
man	Displays the function manual.

Parent topic: Kunpeng AutoTuner