运用训练结果

命令功能

使用已有的调优结果，可对性能结果展示、使用对应轮次的参数复测、对应用参数赋值、参数重置和进行参数重要性分析等。

命令格式

devkit kat use [-h] -i <dir> [-l {0,1,2,3}]

参数说明

表1 参数说明
参数	参数选项	说明
-h/--help	-	可选参数，获取帮助信息。
-l/--log-level	0/1/2/3	可选参数，设置日志级别，默认为2。说明：新增功能采用更合理设计，默认等级调整为2（WARNING）。 0：日志级别为DEBUG。 1：日志级别为INFO。 2：日志级别为WARNING。 3：日志级别为ERROR。
-i/--input	-	必选参数，配置已自动调优后得到的结果目录。

表2 子命令说明
子命令	选项	说明
example	-	展示使用示例。
exit	-	退出交互界面。
help	-	获取帮助信息。
ibesttest	-	使用参数重要性分析推断出的最佳参数进行测试。说明：对最佳参数有疑问时，可执行此命令推断最佳参数进行测试。
iexample	-	展示参数重要性子命令的示例。
iglobal	-	展示全部参数对调优的重要性。说明：多轮次调优时若需了解哪些参数对性能提升有效果，可执行此命令。回显信息中“importance(%)”列的比值越高，对应的参数对于调优的相关性越高。
iround	-	展示第N轮的局部参数重要性。说明：针对某一轮次的性能有疑问时，可执行此命令查看指定轮次的参数对于性能提升的影响。
log	0/1/2/3	设置日志级别，默认为2。 0：日志级别为DEBUG。 1：日志级别为INFO。 2：日志级别为WARNING。 3：日志级别为ERROR。
reset	-	将应用参数和系统参数恢复赋值前。
set	-	将某轮次运行的参数值设置到应用和系统中。说明：适用于用户对某轮次调优性能数据认可的场景。 set是直接改变用户的参数值，请小心操作。
show	-	展示某些轮次的性能测试数据。
test	-	让某些轮次按其参数再次进行测试。
top	-	展示性能指标最好的前N个轮次的数据。
train	-	继续自动调优，任务的参数配置根据业务场景指标自动调优。

可在子命令后使用"r+数字"指定按性能排名的轮次，如：
show r1：展示性能数据第一的轮次的数据

set r1：将性能数据第一的轮次的任务参数设置到应用和系统中

test r1：使用性能数据第一的轮次的参数再次进行性能测试
可在子命令后使用单个数字表示轮次，如：
show 5：展示自动调优第五轮次的数据

set 5：将自动调优第五轮次的任务参数设置到应用和系统中

test 5：使用自动调优第五轮次任务参数再次进行性能测试

表3 train子命令参数说明
参数	说明
-h/--help	可选参数，获取帮助信息。
-r/--round	可选参数，指定训练的轮次，默认为50轮。
--algorithm	可选参数，指定训练时的算法，仅支持HPO。
-i/--initial-random-ratio	可选参数，指定初始随机比率，该值越大，训练时的随机轮次越多，默认值0，取值范围0-1。

使用示例

对已自动调优的结果报告（结果目录可在开启自动调优最后输出回显中获取）查看：

devkit kat use -i /opt/template/template_Postgresql_20250119_073054/train-20250225-112832

返回信息如下：

Available Commands:

    example     - Show examples for use
    exit        - Exit the program
    help        - Show this help message
    ibesttest   - test best parameters inferenced by parameter importance analysis
    iexample    - Show examples for iglobal and iround
    iglobal     - show global parameter importance
    iround      - show local parameter importance of round <n>
    log         - change the log level
    reset       - reset params
    set         - set the config of round n.
    show        - show round <n> information.
    test        - Retest round sequence. test <n> times.
    top         - show top N performance
    train       - Additional training after original result

For more usage, please refer to "help <command>". e.g., help test

查看使用示例。

输入example，按下“Enter”。

返回信息提供一个简单的回显示例：

To understand the use function, see the following example report of automatic
training.

Example report of an automatic training task (including multiple rounds of
performance test results and performance improvement data):

Total round : 10
Baseline performance: 5.0
--------------------------------------------------------------------------------
| Rank |     Round     |  Performance   |  Improvement  (%) |
--------------------------------------------------------------------------------
|   1  |   round    5  |      10.0      |      100.00       |
|   2  |   round    3  |       9.0      |       80.00       |
|   3  |   round    2  |       8.0      |       60.00       |
|   4  |   round    4  |       7.0      |       40.00       |
|   5  |   round    1  |       6.0      |       20.00       |
--------------------------------------------------------------------------------

Usage:
    The show, set, and test subcommands display the data of each round, assign
    a value to the task parameter of each round, and use the task parameter of
    each round for retesting.

    (1) You can add the letter r plus a number to a subcommand to specify which
    round to be ranked by performance. For example, r1 indicates the first row
    of the training result. You can use:

    show r1      # Display the result data of the round whose performance ranks
                 # first.
    set r1       # Assign a value to the task parameter of the first round of
                 # performance task.
    test r1      # Use the parameter of the first round of performance test to
                 # perform the performance test again.

    (2) You can add a number to a subcommand to specify a specific round. For
    example, perform operations on the fifth round result (the first row in
    the training result table):

    show 5       # Display the result data of the fifth round of automatic
                 # training.
    set 5        # Assign a value to the task parameter of the fifth round of
                 # task.
    test 5       # Use the parameter of the fifth round of task to perform the
                 # performance test again.

    (3) When using the test subcommand, you can specify multiple rounds of data
    and use the round parameters to perform the performance test again. Separate
    the rounds using commas (,).
    For example:

    test 1,2,r1,r2    # Use the parameters of the first and second training rounds
                      # and the first and second rounds with the best performance
                      # data to perform the performance test again (corresponding
                      # to rows 1, 2, 5, and 3 in the table).

    You can add top or n-m to the test command to specify multiple rounds.
    For example:
    test 1-3,top3     # Rounds 1, 2, and 3 (corresponding to rows 2, 3, and 5 in
                      # the table) and the three rounds with the best performance
                      # data (corresponding to rows 1, 2, and 3 in the table)

    You can add the specific rounds to the test command and then a space and a
    number after the specific rounds to specify the number of repeated performance
    tests for these rounds. For example:
    test 1,2 3        # Retest rounds: 1,2,1,2,1,2

获取帮助信息。

输入help后按“Enter”；可在help后键入需要查看的子命令获取对应子命令的帮助详情。

Available Commands:

    example     - Show examples for use
    exit        - Exit the program
    help        - Show this help message
    
    iexample    - Show examples for iglobal and iround
    iglobal     - show global parameter importance
    iround      - show local parameter importance of round <n>
    log         - change the log level
    reset       - reset params
    set         - set the config of round n.
    show        - show round <n> information.
    test        - Retest round sequence. test <n> times.
    top         - show top N performance

For more usage, please refer to "help <command>". e.g., help test

查看性能测试数据。

输入show 1后按下“Enter”，可查看自动调优第一轮详细任务参数和性能数据；直接输入show并按下“Enter”，可查看全部轮次性能数据。

返回信息：

[2025-02-25 09:52:11 UTC] [KAT] [message] - ======================== info of Round   1   ===================================
[2025-02-25 09:52:11 UTC] [KAT] [message] - # Performance: 91.07
[2025-02-25 09:52:11 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 09:52:11 UTC] [KAT] [message] - ----------------------          Params         ---------------------------------
[2025-02-25 09:52:11 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 09:52:11 UTC] [KAT] [message] -     Postgresql:
[2025-02-25 09:52:11 UTC] [KAT] [message] -         autovacuum_vacuum_cost_delay                                 = 96
[2025-02-25 09:52:11 UTC] [KAT] [message] -         autovacuum_vacuum_cost_limit                                 = 1994
[2025-02-25 09:52:11 UTC] [KAT] [message] -         autovacuum_vacuum_scale_factor                               = 0.07744139964226633
[2025-02-25 09:52:11 UTC] [KAT] [message] -         autovacuum_vacuum_threshold                                  = 714
...
...
...
[2025-02-25 09:52:11 UTC] [KAT] [message] -         parallel_leader_participation                                = on
[2025-02-25 09:52:11 UTC] [KAT] [message] -         vacuum_cleanup_index_scale_factor                            = 0.8960460592061281
[2025-02-25 09:52:11 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 09:52:11 UTC] [KAT] [message] - ----------------------          Run Info        --------------------------------
[2025-02-25 09:52:11 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 09:52:11 UTC] [KAT] [message] - Run record: 1
[2025-02-25 09:52:11 UTC] [KAT] [message] - Result: success
[2025-02-25 09:52:11 UTC] [KAT] [message] - Performance: 91.07
[2025-02-25 09:52:11 UTC] [KAT] [message] - Run time: 2025-02-25 09:38:24.265750
[2025-02-25 09:52:11 UTC] [KAT] [message] - Finish time: P2025-02-25 09:38:24.265759
[2025-02-25 09:52:11 UTC] [KAT] [message] -
[2025-02-25 09:52:11 UTC] [KAT] [message] - ================================================================================

查看性能前三的数据。

输入top 3后按“Enter”。

返回信息：

[2025-02-25 09:53:05 UTC] [KAT] [message] - ============================== Auto Tuning Report ==============================
[2025-02-25 09:53:05 UTC] [KAT] [message] - Total round             : 5
[2025-02-25 09:53:05 UTC] [KAT] [message] - Total run               : 6
[2025-02-25 09:53:05 UTC] [KAT] [message] - Fail times              : 0
[2025-02-25 09:53:05 UTC] [KAT] [message] - Application             : Postgresql
[2025-02-25 09:53:05 UTC] [KAT] [message] - Application version     : 11.3
[2025-02-25 09:53:05 UTC] [KAT] [message] - Test tool               : benchmarksql
[2025-02-25 09:53:05 UTC] [KAT] [message] - Tuning direction        : high
[2025-02-25 09:53:05 UTC] [KAT] [message] - Algorithm               : HPO
[2025-02-25 09:53:05 UTC] [KAT] [message] - Random initial ratio    : 0.5
[2025-02-25 09:53:05 UTC] [KAT] [message] - Performance description : tpmC, transactions per minute
[2025-02-25 09:53:05 UTC] [KAT] [message] - Parameter groups        : Postgresql
[2025-02-25 09:53:05 UTC] [KAT] [message] - Baseline performance    : 97.03
[2025-02-25 09:53:05 UTC] [KAT] [message] - Top 3 performance :
[2025-02-25 09:53:05 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 09:53:05 UTC] [KAT] [message] - |  Rank  |     Round     |  Performance   |  Improvement  (%) |
[2025-02-25 09:53:05 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 09:53:05 UTC] [KAT] [message] - |   1    |   round  1    |     91.07      |       -6.14       |
[2025-02-25 09:53:05 UTC] [KAT] [message] - |   2    |   round  4    |     86.92      |      -10.42       |
[2025-02-25 09:53:05 UTC] [KAT] [message] - |   3    |   round  3    |     79.44      |      -18.13       |
[2025-02-25 09:53:05 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 09:53:05 UTC] [KAT] [message] - Note:
[2025-02-25 09:53:05 UTC] [KAT] [message] - The performance value is the return value of the run test step,
[2025-02-25 09:53:05 UTC] [KAT] [message] - Performance improvement (%) =
[2025-02-25 09:53:05 UTC] [KAT] [message] -     (round performance - baseline performance) / baseline performance * 100
[2025-02-25 09:53:05 UTC] [KAT] [message] - Case package locate: /opt/template/template_Postgresql_20250119_073054/train-20250225-112832
[2025-02-25 09:53:05 UTC] [KAT] [message] - ================================================================================

对指定轮次再次进行性能测试。

输入test r1后按“Enter”。

在使用参数赋值前，为确定调优参数是否稳定，可对性能提升较好的轮次进行多次复测，结果中变异系数“CV(%)”越小，则表明复测的性能数据一致性越高，对应轮次的调优参数越稳定，可酌情使用到应用中。

合理的测试场景下，变异系数“CV(%)”低于1%认为参数对应优化的效果非常稳定，1%-5%是比较稳定，5%-10%不太稳定可根据实际情况是否接收，若超过10%则认为此轮的参数不稳定，建议优先解决性能数据的稳定性问题（环境或人为因素）。

使用子命令“test”时，可指定多个轮次数据并使用对应轮次的参数再次性能测试，轮次间使用“,”分隔，如：
test 1,2,r1,r2：使用调优第一、第二轮以及性能数据第一、第二的轮次参数，再次进行性能测试。
在“test”命令后使用top或“n-m”可指定多个轮次：
test 1-3,top3：自动调优的第1、2、3轮和性能数据最好的前三轮。
在“test”命令指定轮次后键入“空格+数字”可对其指定轮设置重复性能测试次数，如：
test 1,2 3：对自动调优第一轮和第二轮复测循环三次。
在一般情况下，通常使用AB轮的方式进行复测，来保证测试结果的稳定性。例如：需要对最高性能轮次（假设轮次号为A）的结果进行复测，可以使用test 0,A 5，对基线轮次和最高性能轮次交替测试5组，共10次。

返回信息：

[2025-02-25 09:53:48 UTC] [KAT] [message] - ================================================================================
[2025-02-25 09:53:48 UTC] [KAT] [message] -           kunpeng automatic tuning - task set up
[2025-02-25 09:53:48 UTC] [KAT] [message] -           start time: 2025-02-25 09:53:48
[2025-02-25 09:53:48 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 09:53:51 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step   <params prepare>   start -----------
[2025-02-25 09:53:51 UTC] [KAT] [message] - ================================================================================
[2025-02-25 09:53:51 UTC] [KAT] [message] -           kunpeng automatic tuning - test 1
[2025-02-25 09:53:51 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step    <round set up>    start -----------
[2025-02-25 09:53:51 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step   <assign params>    start -----------
[2025-02-25 09:53:53 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step      <run test>      start -----------
[2025-02-25 09:54:01 UTC] [KAT] [message] - --------- Run test successfully.
[2025-02-25 09:54:01 UTC] [KAT] [message] - --------- Performance     :       99.85
[2025-02-25 09:54:01 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step  <round tear down>   start -----------
[2025-02-25 09:54:02 UTC] [KAT] [message] - ================================ Detail Report =================================
[2025-02-25 09:54:02 UTC] [KAT] [message] - Actual test times : 1
[2025-02-25 09:54:02 UTC] [KAT] [message] - Fail times        : 0
[2025-02-25 09:54:02 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 09:54:02 UTC] [KAT] [message] - |  No.  |  Round  | Status | Performance | Performance(before) |
[2025-02-25 09:54:02 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 09:54:02 UTC] [KAT] [message] - |   1   |    1    |   S    |    99.85    |        91.07        |
[2025-02-25 09:54:02 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 09:54:02 UTC] [KAT] [message] - ================================  Total Report =================================
[2025-02-25 09:54:02 UTC] [KAT] [message] - Tested rounds     : 1
[2025-02-25 09:54:02 UTC] [KAT] [message] - Tuning Direction : high
[2025-02-25 09:54:02 UTC] [KAT] [message] - Performance description: tpmC, transactions per minute
[2025-02-25 09:54:02 UTC] [KAT] [message] - Baseline performance  : 97.03
[2025-02-25 09:54:02 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 09:54:02 UTC] [KAT] [message] - |  Round  | Times | Performance | Performance(before) | Change (%) |  CV(%) |
[2025-02-25 09:54:02 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 09:54:02 UTC] [KAT] [message] - |    1    |   1   |    99.85    |        91.07        |   +9.64    |  0.00  |
[2025-02-25 09:54:02 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 09:54:02 UTC] [KAT] [message] - ================================================================================

调整输出日志等级。
输入log 1后按下“Enter”，可将回显打印信息等级修改为“INFO”级，仅当次交互界面有效。

查看全部参数对调优的重要性。

输入iglobal后按“Enter”。

importance(%)列的百分比值越高，表明该行参数对于调优的相关性越高（对性能影响越大），可重点关注。

返回信息：

   Param Group     |     Parameter      | importance(%) |       range        |            Top 5
-------------------------------------------------------------------------------------------------------------
rocksdb             |max_background_flush|12.79%         |[1, 64]             |['31', '18', '7', '18', '35']
                    |es                  |               |                    |
-------------------------------------------------------------------------------------------------------------
rocksdb             |target_file_size_bas|10.79%         |[2, 268435456]      |['3144477', '18231017', '58331
                    |e                   |               |                    |577', '18231017', '123520730']
-------------------------------------------------------------------------------------------------------------
rocksdb             |subcompactions      |10.54%         |[1, 32]             |['14', '22', '26', '26', '19']
-------------------------------------------------------------------------------------------------------------
...
...
...
-------------------------------------------------------------------------------------------------------------
rocksdb             |pin_l0_filter_and_in|0.31%          |('true', 'false')   |['false', 'false', 'false', 'f
                    |dex_blocks_in_cache |               |                    |alse', 'false']
-------------------------------------------------------------------------------------------------------------

查看某一轮调优的参数重要性。

输入iround 1后按“Enter”。

Baseline performance：实际测试得到的基线性能。

Round N performance：轮次N实际测试得到的性能。

Expected performance：模型基于训练历史计算得到的，当所有参数的收益都为零时的期望性能。

Round N estimated performance：模型基于训练历史计算得到的轮次N预测性能。

Estimated total benefit：轮次N预测性能和期望性能的差值，表示所有参数获取到的总收益。

返回信息：

Baseline performance: 19516955.00
Round 1 performance: 19878245.0
Expected performance: 19747731.20
Round 1 estimated performance: 19836562.31
Estimated total benefit: 88831.11

    Param Group     |     Parameter      |  Param Value  | Estimated Benefit
------------------------------------------------------------------------------
rocksdb             |max_background_flush|7              |19063.97
                    |es                  |               |
------------------------------------------------------------------------------
rocksdb             |subcompactions      |26             |16060.71
------------------------------------------------------------------------------
rocksdb             |target_file_size_bas|18231017       |15296.90
                    |e                   |               |
------------------------------------------------------------------------------
...
...
...
------------------------------------------------------------------------------
rocksdb             |pin_l0_filter_and_in|false          |371.09
                    |dex_blocks_in_cache |               |
------------------------------------------------------------------------------
rocksdb             |delayed_write_rate  |33446408       |112.15
------------------------------------------------------------------------------

使用参数重要性分析推断出的最佳参数进行测试。

输入ibesttest后按“Enter”。

返回信息如下：

Baseline performance: 24592683.0
Best parameters inferenced:
  Param Group  |          Parameter           |    Param Value
-------------------------------------------------------------------
rocksdb        |batch_size                    |109
-------------------------------------------------------------------
rocksdb        |block_size                    |142561
-------------------------------------------------------------------
rocksdb        |bloom_bits                    |77
-------------------------------------------------------------------
...
...
...
-------------------------------------------------------------------
rocksdb        |target_file_size_base         |34462869
-------------------------------------------------------------------
rocksdb        |write_buffer_size             |21101528
-------------------------------------------------------------------
[2025-02-18 07:24:41 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step    <round set up>    start -----------------
[2025-02-18 07:24:41 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step   <assign params>    start -----------------
[2025-02-18 07:24:42 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step      <run test>      start -----------------
[2025-02-18 07:24:42 UTC] [KAT] [message] -  ### Run test successfully. Performance of this round is [28985507] ###
[2025-02-18 07:24:42 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step  <round tear down>   start -----------------

继续调优。

输入train -r 2后按“Enter”，可在use界面中继续两轮的自动调优。

最后的报告将和之前自动调优数据一同汇总排序显示。

返回信息：

[2025-02-25 11:12:21 UTC] [KAT] [message] - Algorithm               : HPO
[2025-02-25 11:12:21 UTC] [KAT] [message] - ================================================================================
[2025-02-25 11:12:21 UTC] [KAT] [message] -           kunpeng automatic tuning - round 6
[2025-02-25 11:12:21 UTC] [KAT] [message] -           start time: 2025-02-25 11:12:21
[2025-02-25 11:12:21 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 11:12:21 UTC] [KAT] [message] - Due to the presence of historical data and the random iterations being set to 0, the inference tuning phase begins.
[2025-02-25 11:12:33 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step    <round set up>    start -----------
[2025-02-25 11:12:33 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step   <assign params>    start -----------
[2025-02-25 11:12:35 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step      <run test>      start -----------
[2025-02-25 11:12:42 UTC] [KAT] [message] - --------- Run test successfully.
[2025-02-25 11:12:42 UTC] [KAT] [message] - --------- Performance     :       83.74
[2025-02-25 11:12:42 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step  <round tear down>   start -----------
[2025-02-25 11:12:42 UTC] [KAT] [message] - ================================================================================
[2025-02-25 11:12:42 UTC] [KAT] [message] -           kunpeng automatic tuning - round 7
[2025-02-25 11:12:42 UTC] [KAT] [message] -           start time: 2025-02-25 11:12:42
[2025-02-25 11:12:42 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 11:12:42 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step    <round set up>    start -----------
[2025-02-25 11:12:42 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step   <assign params>    start -----------
[2025-02-25 11:12:44 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step      <run test>      start -----------
[2025-02-25 11:12:52 UTC] [KAT] [message] - --------- Run test successfully.
[2025-02-25 11:12:52 UTC] [KAT] [message] - --------- Performance     :       81.41
[2025-02-25 11:12:52 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step  <round tear down>   start -----------
[2025-02-25 11:12:52 UTC] [KAT] [message] - ============================== Auto Tuning Report ==============================
[2025-02-25 11:12:52 UTC] [KAT] [message] - Total round             : 7
[2025-02-25 11:12:52 UTC] [KAT] [message] - Total run               : 12
[2025-02-25 11:12:52 UTC] [KAT] [message] - Fail times              : 0
[2025-02-25 11:12:52 UTC] [KAT] [message] - Application             : Postgresql
[2025-02-25 11:12:52 UTC] [KAT] [message] - Application version     : 11.3
[2025-02-25 11:12:52 UTC] [KAT] [message] - Test tool               : benchmarksql
[2025-02-25 11:12:52 UTC] [KAT] [message] - Tuning direction        : high
[2025-02-25 11:12:52 UTC] [KAT] [message] - Algorithm               : HPO
[2025-02-25 11:12:52 UTC] [KAT] [message] - Random initial ratio    : 0.5
[2025-02-25 11:12:52 UTC] [KAT] [message] - Performance description : tpmC, transactions per minute
[2025-02-25 11:12:52 UTC] [KAT] [message] - Parameter groups        : Postgresql
[2025-02-25 11:12:52 UTC] [KAT] [message] - Baseline performance    : 97.03
[2025-02-25 11:12:52 UTC] [KAT] [message] - Top 10 performance :
[2025-02-25 11:12:52 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 11:12:52 UTC] [KAT] [message] - |  Rank  |     Round     |  Performance   |  Improvement  (%) |
[2025-02-25 11:12:52 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 11:12:52 UTC] [KAT] [message] - |   1    |   round  4    |     86.92      |      -10.42       |
[2025-02-25 11:12:52 UTC] [KAT] [message] - |   2    |   round  6    |     83.74      |      -13.70       |
[2025-02-25 11:12:52 UTC] [KAT] [message] - |   3    |   round  7    |     81.41      |      -16.10       |
[2025-02-25 11:12:52 UTC] [KAT] [message] - |   4    |   round  1    |     80.32      |      -17.22       |
[2025-02-25 11:12:52 UTC] [KAT] [message] - |   5    |   round  3    |     79.44      |      -18.13       |
[2025-02-25 11:12:52 UTC] [KAT] [message] - |   6    |   round  5    |     70.05      |      -27.81       |
[2025-02-25 11:12:52 UTC] [KAT] [message] - |   7    |   round  2    |     66.37      |      -31.60       |
[2025-02-25 11:12:52 UTC] [KAT] [message] - --------------------------------------------------------------------------------
[2025-02-25 11:12:52 UTC] [KAT] [message] - Note:
[2025-02-25 11:12:52 UTC] [KAT] [message] - The performance value is the return value of the run test step,
[2025-02-25 11:12:52 UTC] [KAT] [message] - Performance improvement (%) =
[2025-02-25 11:12:52 UTC] [KAT] [message] -     (round performance - baseline performance) / baseline performance * 100
[2025-02-25 11:12:52 UTC] [KAT] [message] - Case package locate: /opt/template/template_Postgresql_20250119_073054/train-20250225-112832
[2025-02-25 11:12:52 UTC] [KAT] [message] - ================================================================================

若对性能第一轮次的调优性能数据认可，可将性能第一的自动调优轮次的参数设置到应用和系统中。
输入set r1后按下“Enter”。

参数设置成功的返回信息：
1
[2025-02-25 11:14:31 UTC] [KAT] [message] - set params of r1 success!

重置任务参数到训练调优前。

输入reset后按“Enter”。

重置成功时返回信息：

[2025-02-25 11:14:55 UTC] [KAT] [message] - --------- kunpeng automatic tuning - Step   <params recover>   start -----------

退出交互界面。
输入exit，按下“Enter”。

父主题： 鲲鹏自动调优