东京热久久综合久久88,美女熟女一二区不卡,欧美成人精品网站播放

PolarDB MySQL版新增列存索引查詢性能分析功能。該功能通過(guò)采集執(zhí)行SQL語(yǔ)句時(shí)的耗時(shí)等信息，并結(jié)合EXPLAIN獲得的查詢計(jì)劃一并返回給用戶。能夠幫助用戶了解執(zhí)行SQL語(yǔ)句時(shí)的耗時(shí)細(xì)節(jié)，并且能夠輔助分析慢SQL。

版本要求

集群版本需為PolarDB MySQL版8.0.1.1.42及以上，您可以通過(guò)查詢版本號(hào)來(lái)確認(rèn)集群版本。

使用說(shuō)明

您需要將參數(shù)imci_analyze_query的值設(shè)置為ON，來(lái)開(kāi)啟列存索引查詢性能分析功能。

參數(shù)名稱

級(jí)別

說(shuō)明

imci_analyze_query

Session

列存索引查詢性能分析功能控制開(kāi)關(guān)。取值范圍如下：

OFF（默認(rèn)）：關(guān)閉列存索引查詢性能分析功能。
ON：開(kāi)啟列存索引查詢性能分析功能。

示例

簡(jiǎn)單查詢示例

以在TPC-H Schema上執(zhí)行的一個(gè)簡(jiǎn)單查詢語(yǔ)句為例，SQL語(yǔ)句如下：

SELECT l_shipmode, COUNT(*) FROM lineitem, orders WHERE l_orderkey = o_orderkey GROUP BY l_shipmode;

首先，您可以通過(guò)EXPLAIN得到該查詢優(yōu)化器估計(jì)的各個(gè)算子的執(zhí)行代價(jià)以及返回行數(shù)信息。

EXPLAIN SELECT l_shipmode, COUNT(*) FROM lineitem, orders WHERE l_orderkey = o_orderkey GROUP BY l_shipmode;

查詢結(jié)果如下：

+----+------------------------+----------+--------+----------+---------------------------------------------------------------+
| ID | Operator               | Name     | E-Rows | E-Cost   | Extra Info                                                    |
+----+------------------------+----------+--------+----------+---------------------------------------------------------------+
|  1 | Select Statement       |          |        |          | IMCI Execution Plan (max_dop = 1, max_query_mem = unlimited)  |
|  2 | └─Hash Groupby         |          | 6      | 51218.50 | Group Key: lineitem.l_shipmode                                |
|  3 |   └─Compute Scalar     |          | 1869   | 50191.00 |                                                               |
|  4 |     └─Hash Join        |          | 1869   | 31000.00 | Join Cond: lineitem.l_orderkey = orders.o_orderkey            |
|  5 |       ├─Table Scan     | lineitem | 2000   | 0.00     |                                                               |
|  6 |       └─Table Scan     | orders   | 2000   | 0.00     |                                                               |
+----+------------------------+----------+--------+----------+---------------------------------------------------------------+

開(kāi)啟性能分析功能。
```
SET imci_analyze_query = ON;
```
執(zhí)行SQL語(yǔ)句。
```
SELECT l_shipmode, COUNT(*) FROM lineitem, orders WHERE l_orderkey = o_orderkey GROUP BY l_shipmode;
```
執(zhí)行結(jié)果如下：
```
+------------+----------+
| l_shipmode | COUNT(*) |
+------------+----------+
| REG AIR    |      283 |
| SHIP       |      269 |
| FOB        |      299 |
| RAIL       |      289 |
| TRUCK      |      314 |
| MAIL       |      274 |
| AIR        |      272 |
+------------+----------+
7 rows in set (0.05 sec)
```
性能分析結(jié)果會(huì)存放在information_schema.imci_sql_profiling表中。在將參數(shù)imci_analyze_query的值設(shè)置為ON時(shí)，該表將存放最近一條使用列存索引執(zhí)行的查詢語(yǔ)句的性能分析信息，通過(guò)查詢這張表，就可以獲得執(zhí)行的查詢語(yǔ)句的性能信息。

查詢SQL語(yǔ)句的性能信息。

/*ROUTE_TO_LAST_USED*/SELECT * FROM information_schema.imci_sql_profiling;

說(shuō)明

/*ROUTE_TO_LAST_USED*/用于指示智能代理將查詢路由到上一條語(yǔ)句執(zhí)行的節(jié)點(diǎn)上。否則可能會(huì)因?yàn)槁酚刹坏讲樵儓?zhí)行的節(jié)點(diǎn)，導(dǎo)致查詢不到結(jié)果或結(jié)果不正確。

查詢結(jié)果如下：

+----+------------------------+----------+--------+----------+-------------------+--------------------------------------------------------------------------------------------------------+
| ID | Operator               | Name     | A-Rows | A-Cost   | Execution Time(s) | Extra Info                                                                                             |
+----+------------------------+----------+--------+----------+-------------------+--------------------------------------------------------------------------------------------------------+
|  1 | Select Statement       |          |      0 | 52609.51 |                 0 | IMCI Execution Plan (max_dop = 1, real_dop = 1, max_query_mem = unlimited, real_query_mem = unlimited) |
|  2 | └─Hash Groupby         |          |      7 |  52609.5 |             0.002 | Group Key: lineitem.l_shipmode                                                                         |
|  3 |   └─Compute Scalar     |          |   2000 |    51501 |                 0 |                                                                                                        |
|  4 |     └─Hash Join        |          |   2000 |    31000 |             0.007 | Join Cond: lineitem.l_orderkey = orders.o_orderkey                                                     |
|  5 |       ├─Table Scan     | lineitem |   2000 |        0 |             0.001 |                                                                                                        |
|  6 |       └─Table Scan     | orders   |   2000 |        0 |                 0 |                                                                                                        |
+----+------------------------+----------+--------+----------+-------------------+--------------------------------------------------------------------------------------------------------+

可以看到，Hash Join與Hash Groupby對(duì)應(yīng)的Extra Info信息中標(biāo)明了對(duì)應(yīng)的連接條件以及分組列，而第1行的Extra Info則展示了執(zhí)行查詢操作時(shí)使用的CPU和內(nèi)存信息。

（可選）您可以在information_schema.imci_sql_profiling表上執(zhí)行聚合等操作。如查看該查詢語(yǔ)句的執(zhí)行時(shí)間等。
```
/*ROUTE_TO_LAST_USED*/SELECT SUM(`Execution Time(s)`) AS TOTAL_TIME FROM information_schema.imci_sql_profiling;
```
執(zhí)行結(jié)果如下：
```
+----------------------+
| TOTAL_TIME           |
+----------------------+
| 0.010000000000000002 |
+----------------------+
1 row in set (0.00 sec)
```
可以看到，該查詢語(yǔ)句的執(zhí)行耗時(shí)為10ms。

復(fù)雜查詢示例

以在TPC-H SF100數(shù)據(jù)集上執(zhí)行一個(gè)復(fù)雜查詢?yōu)槔ㄟ^(guò)EXPLAIN和查詢分析功能分析查詢耗時(shí)分布以及性能。SQL語(yǔ)句如下：

SELECT
    c_name,
    sum(l_quantity)
FROM
    customer,
    orders,
    lineitem
WHERE
    o_orderkey IN (
        SELECT
            l_orderkey
        FROM
            lineitem
        WHERE 
            l_partkey > 18000000
    )
    AND c_custkey = o_custkey
    AND o_orderkey = l_orderkey
GROUP BY
    c_name
ORDER BY
    c_name
LIMIT 10;

通過(guò)EXPLAIN查看該查詢語(yǔ)句的執(zhí)行計(jì)劃。

EXPLAIN SELECT
    c_name,
    sum(l_quantity)
FROM
    customer,
    orders,
    lineitem
WHERE
    o_orderkey IN (
        SELECT
            l_orderkey
        FROM
            lineitem
        WHERE 
            l_partkey > 18000000
    )
    AND c_custkey = o_custkey
    AND o_orderkey = l_orderkey
GROUP BY
    c_name
ORDER BY
    c_name
LIMIT 10;

執(zhí)行結(jié)果如下：

+----+----------------------------------+----------+-----------+------------+---------------------------------------------------------------+
| ID | Operator                         | Name     | E-Rows    | E-Cost     | Extra Info                                                    |
+----+----------------------------------+----------+-----------+------------+---------------------------------------------------------------+
|  1 | Select Statement                 |          |           |            | IMCI Execution Plan (max_dop = 32, max_query_mem = unlimited) |
|  2 | └─Limit                          |          | 10        | 7935739.96 | Offset=0 Limit=10                                             |
|  3 |   └─Sort                         |          | 10        | 7935739.96 | Sort Key: c_name ASC                                          |
|  4 |     └─Hash Groupby               |          | 1503700   | 7933273.99 | Group Key: customer.C_NAME                                    |
|  5 |       └─Hash Right Semi Join     |          | 54010545  | 7865930.26 | Join Cond: lineitem.L_ORDERKEY = orders.O_ORDERKEY            |
|  6 |         ├─Table Scan             | lineitem | 59785766  | 24001.52   | Cond: (L_PARTKEY > 18000000)                                  |
|  7 |         └─Hash Join              |          | 538776190 | 5488090.33 | Join Cond: orders.O_ORDERKEY = lineitem.L_ORDERKEY            |
|  8 |           ├─Hash Join            |          | 181006430 | 668535.99  | Join Cond: customer.C_CUSTKEY = orders.O_CUSTKEY              |
|  9 |           │ ├─Table Scan         | customer | 15000000  | 600.00     |                                                               |
| 10 |           │ └─Table Scan         | orders   | 150000000 | 6000.00    |                                                               |
| 11 |           └─Table Scan           | lineitem | 600037902 | 24001.52   |                                                               |
+----+----------------------------------+----------+-----------+------------+---------------------------------------------------------------+

該SQL語(yǔ)句的執(zhí)行代價(jià)為7935739。其中，ID為7的Hash Join會(huì)消耗大量CPU（其代價(jià)為5488090，占總代價(jià)約70%）。若要驗(yàn)證這條查詢語(yǔ)句的性能問(wèn)題，您可以開(kāi)啟性能分析功能，開(kāi)啟后再執(zhí)行該SQL語(yǔ)句，并通過(guò)查看SQL語(yǔ)句的查詢性能信息來(lái)分析具體原因。

開(kāi)啟性能分析功能。
```
SET imci_analyze_query = ON;
```

執(zhí)行SQL語(yǔ)句。

SELECT
    c_name,
    sum(l_quantity)
FROM
    customer,
    orders,
    lineitem
WHERE
    o_orderkey IN (
        SELECT
            l_orderkey
        FROM
            lineitem
        WHERE 
            l_partkey > 18000000
    )
    AND c_custkey = o_custkey
    AND o_orderkey = l_orderkey
GROUP BY
    c_name
ORDER BY
    c_name
LIMIT 10;

執(zhí)行結(jié)果如下：

+--------------------+-----------------+
| c_name             | sum(l_quantity) |
+--------------------+-----------------+
| Customer#000000001 |          172.00 |
| Customer#000000002 |          663.00 |
| Customer#000000004 |          174.00 |
| Customer#000000005 |          488.00 |
| Customer#000000007 |         1135.00 |
| Customer#000000008 |          440.00 |
| Customer#000000010 |          625.00 |
| Customer#000000011 |          143.00 |
| Customer#000000013 |         1032.00 |
| Customer#000000014 |          564.00 |
+--------------------+-----------------+
10 rows in set (21.37 sec)

查詢SQL語(yǔ)句的性能信息。

/*ROUTE_TO_LAST_USED*/SELECT * FROM information_schema.imci_sql_profiling;

說(shuō)明

/*ROUTE_TO_LAST_USED*/用于指示智能代理將查詢路由到上一條語(yǔ)句執(zhí)行的節(jié)點(diǎn)上。否則可能會(huì)因?yàn)槁酚傻藉e(cuò)誤節(jié)點(diǎn)而查詢不到結(jié)果。

查詢結(jié)果如下：

+----+----------------------------------+----------+-----------+------------+-------------------+----------------------------------------------------+
| ID | Operator                         | Name     | A-Rows    | A-Cost     | Execution Time(s) | Extra Info                                         |
+----+----------------------------------+----------+-----------+------------+-------------------+----------------------------------------------------+
|  1 | Select Statement                 |          |         0 | 8336856.67 |                 0 |                                                    |
|  2 | └─Limit                          |          |        10 | 8336856.67 |             0.002 | Offset=0 Limit=10                                  |
|  3 |   └─Sort                         |          |         0 | 8336856.67 |             2.275 | Sort Key: c_name ASC                               |
|  4 |     └─Hash Groupby               |          |   9813586 | 8320763.22 |           160.083 | Group Key: customer.C_NAME                         |
|  5 |       └─Hash Right Semi Join     |          | 239598134 | 7994854.23 |            98.174 | Join Cond: lineitem.L_ORDERKEY = orders.O_ORDERKEY |
|  6 |         ├─Table Scan             | lineitem |  60013756 |   24001.52 |              3.28 | Cond: (L_PARTKEY > 18000000)                       |
|  7 |         └─Hash Join              |          | 600037902 | 5156677.35 |           301.503 | Join Cond: orders.O_ORDERKEY = lineitem.L_ORDERKEY |
|  8 |           ├─Hash Join            |          | 150000000 |  629777.96 |            97.201 | Join Cond: customer.C_CUSTKEY = orders.O_CUSTKEY   |
|  9 |           │ ├─Table Scan         | customer |  15000000 |        600 |             3.321 |                                                    |
| 10 |           │ └─Table Scan         | orders   | 150000000 |       6000 |             0.241 |                                                    |
| 11 |           └─Table Scan           | lineitem | 600037902 |   24001.52 |             0.661 |                                                    |
+----+----------------------------------+----------+-----------+------------+-------------------+----------------------------------------------------+

根據(jù)查詢的性能信息可以看出：Hash Join(7)耗費(fèi)了約一半的時(shí)間。這是因?yàn)?code data-tag="code" code-type="xCode" class="code">lineitem表與Hash Join(8)表的結(jié)果都很大，從而導(dǎo)致join的數(shù)據(jù)量很大，因此耗時(shí)很長(zhǎng)。在Hash Join中，構(gòu)建哈希表的開(kāi)銷通常遠(yuǎn)遠(yuǎn)大于哈希查找，因此只需要減小Join(8)的結(jié)果大小，就能有效減少查詢的耗時(shí)。

查看SQL語(yǔ)句和執(zhí)行計(jì)劃，若優(yōu)化器將o_orderkey in (...)轉(zhuǎn)換為半連接子查詢，且能夠?qū)⒃摋l件下推，則能提高該SQL語(yǔ)句的查詢性能。

開(kāi)啟基于代價(jià)的半連接下推功能。
```
SET imci_optimizer_switch = 'semijoin_pushdown=on';
```

通過(guò)EXPLAIN查看該查詢語(yǔ)句的執(zhí)行計(jì)劃。

EXPLAIN SELECT
    c_name,
    sum(l_quantity)
FROM
    customer,
    orders,
    lineitem
WHERE
    o_orderkey IN (
        SELECT
            l_orderkey
        FROM
            lineitem
        WHERE 
            l_partkey > 18000000
    )
    AND c_custkey = o_custkey
    AND o_orderkey = l_orderkey
GROUP BY
    c_name
ORDER BY
    c_name
LIMIT 10;

執(zhí)行結(jié)果如下：

+----+----------------------------------------+----------+-----------+------------+---------------------------------------------------------------+
| ID | Operator                               | Name     | E-Rows    | E-Cost     | Extra Info                                                    |
+----+----------------------------------------+----------+-----------+------------+---------------------------------------------------------------+
|  1 | Select Statement                       |          |           |            | IMCI Execution Plan (max_dop = 32, max_query_mem = unlimited) |
|  2 | └─Limit                                |          | 10        | 2800433.74 | Offset=0 Limit=10                                             |
|  3 |   └─Sort                               |          | 10        | 2800433.74 | Sort Key: c_name ASC                                          |
|  4 |     └─Hash Groupby                     |          | 14567321  | 2776544.58 | Group Key: customer.C_NAME                                    |
|  5 |       └─Hash Join                      |          | 57918330  | 2631846.75 | Join Cond: orders.O_ORDERKEY = lineitem.L_ORDERKEY            |
|  6 |         ├─Hash Join                    |          | 14567321  | 1014041.92 | Join Cond: orders.O_CUSTKEY = customer.C_CUSTKEY              |
|  7 |         │ ├─Hash Right Semi Join       |          | 12071937  | 906226.67  | Join Cond: lineitem.L_ORDERKEY = orders.O_ORDERKEY            |
|  8 |         │ │ ├─Table Scan               | lineitem | 59785766  | 24001.52   | Cond: (L_PARTKEY > 18000000)                                  |
|  9 |         │ │ └─Table Scan               | orders   | 150000000 | 6000.00    |                                                               |
| 10 |         │ └─Table Scan                 | customer | 15000000  | 600.00     |                                                               |
| 11 |         └─Table Scan                   | lineitem | 600037902 | 24001.52   |                                                               |
+----+----------------------------------------+----------+-----------+------------+---------------------------------------------------------------+

可以看出，開(kāi)啟半連接下推功能后，查詢總代價(jià)從7935739下降至2800433。

重新執(zhí)行該SQL語(yǔ)句。

SELECT
    c_name,
    sum(l_quantity)
FROM
    customer,
    orders,
    lineitem
WHERE
    o_orderkey IN (
        SELECT
            l_orderkey
        FROM
            lineitem
        WHERE 
            l_partkey > 18000000
    )
    AND c_custkey = o_custkey
    AND o_orderkey = l_orderkey
GROUP BY
    c_name
ORDER BY
    c_name
LIMIT 10;

執(zhí)行結(jié)果如下：

+--------------------+-----------------+
| c_name             | sum(l_quantity) |
+--------------------+-----------------+
| Customer#000000001 |          172.00 |
| Customer#000000002 |          663.00 |
| Customer#000000004 |          174.00 |
| Customer#000000005 |          488.00 |
| Customer#000000007 |         1135.00 |
| Customer#000000008 |          440.00 |
| Customer#000000010 |          625.00 |
| Customer#000000011 |          143.00 |
| Customer#000000013 |         1032.00 |
| Customer#000000014 |          564.00 |
+--------------------+-----------------+
10 rows in set (13.74 sec)

可以看出，執(zhí)行耗時(shí)降低了約40%。

查詢SQL語(yǔ)句的性能信息。

/*ROUTE_TO_LAST_USED*/SELECT * FROM information_schema.imci_sql_profiling;

執(zhí)行結(jié)果如下：

+----+----------------------------------------+----------+-----------+------------+-------------------+----------------------------------------------------+
| ID | Operator                               | Name     | A-Rows    | A-Cost     | Execution Time(s) | Extra Info                                         |
+----+----------------------------------------+----------+-----------+------------+-------------------+----------------------------------------------------+
|  1 | Select Statement                       |          |         0 | 4318488.35 |                 0 |                                                    |
|  2 | └─Limit                                |          |        10 | 4318488.35 |             0.002 | Offset=0 Limit=10                                  |
|  3 |   └─Sort                               |          |         0 | 4318488.34 |             3.076 | Sort Key: c_name ASC                               |
|  4 |     └─Hash Groupby                     |          |   9813586 | 4302394.89 |           163.149 | Group Key: customer.C_NAME                         |
|  5 |       └─Hash Join                      |          | 239598134 | 3976485.91 |           151.253 | Join Cond: orders.O_ORDERKEY = lineitem.L_ORDERKEY |
|  6 |         ├─Hash Join                    |          |  49393149 | 1321335.54 |            55.392 | Join Cond: orders.O_CUSTKEY = customer.C_CUSTKEY   |
|  7 |         │ ├─Hash Right Semi Join       |          |  49393149 |  954805.16 |            52.552 | Join Cond: lineitem.L_ORDERKEY = orders.O_ORDERKEY |
|  8 |         │ │ ├─Table Scan               | lineitem |  60013756 |   24001.52 |             2.791 | Cond: (L_PARTKEY > 18000000)                       |
|  9 |         │ │ └─Table Scan               | orders   | 150000000 |       6000 |             0.152 |                                                    |
| 10 |         │ └─Table Scan                 | customer |  15000000 |        600 |             0.028 |                                                    |
| 11 |         └─Table Scan                   | lineitem | 600037902 |   24001.52 |             0.642 |                                                    |
+----+----------------------------------------+----------+-----------+------------+-------------------+----------------------------------------------------+

從性能分析結(jié)果可以看出，半連接被下推，之前的大表Join被消除，查詢耗時(shí)顯著減少。

日本熟妇hd丰满老熟妇,中文字幕一区二区三区在线不卡 ,亚洲成片在线观看,免费女同在线一区二区

分析列存索引查詢性能

版本要求

使用說(shuō)明

示例

簡(jiǎn)單查詢示例

復(fù)雜查詢示例