Question

我有一个突然变得太慢的观点，我不知道如何优化它。这些表目前分别包含15000（@dispatchPallet）和135000（@pickLog）行。

我已经编写了一段最小化的代码来展示下面的重要部分。

DECLARE @dispatchPallet TABLE 
(
    [PICK_PALL_NUM] [bigint] NOT NULL,
    [PALLET_PLACEMENT] [nvarchar](4) NOT NULL,
    [SHIPMENT_ID] [nvarchar](255) NULL
)

DECLARE @pickLog TABLE 
(
    [LINE_NUM] [int] NOT NULL,
    [QTY_PRE] [numeric](9, 2) NULL,
    [QTY_SUF] [numeric](9, 2) NULL,
    [PICK_PALL_NUM] [bigint] NULL,
    [ROWID] [uniqueidentifier] NOT NULL,
    [WEIGHT_GROSS] [numeric](9, 3) NULL,
    [VOLUME] [numeric](9, 3) NULL
)


INSERT INTO @dispatchPallet ([PICK_PALL_NUM], [PALLET_PLACEMENT], [SHIPMENT_ID])
VALUES 
(4797753, 'B', 'SHIPMENT-1'),
(4797752, 'B', 'SHIPMENT-2'),
(4797750, 'B', 'SHIPMENT-3'),
(4797749, 'B', 'SHIPMENT-4'),
(4797739, 'B', 'SHIPMENT-5'),
(4797732, 'B', 'SHIPMENT-6'),
(4797731, 'B', 'SHIPMENT-7'),
(4797730, 'B', 'SHIPMENT-7'),
(4797723, 'B', 'SHIPMENT-8'),
(4797713, 'B', 'SHIPMENT-9')

INSERT INTO @pickLog ([LINE_NUM], [QTY_PRE], [QTY_SUF], [PICK_PALL_NUM], [ROWID], [WEIGHT_GROSS])
VALUES 
(30, 54, 54, 4797753, NEWID(), 1070.280),
(10, 24, 24, 4797752, NEWID(), 471.360),
(30, 12, 12, 4797750, NEWID(), 237.960),
(320, 25, 25, 4797749, NEWID(), 102.750),
(110, 3, 3, 4797739, NEWID(), 40.650),
(40, 12, 12, 4797732, NEWID(), 238.080),
(50, 4, 4, 4797732, NEWID(), 78.560),
(20, 20, 20, 4797731, NEWID(), 110.000),
(20, 40, 40, 4797730, NEWID(), 220.000),
(1340, 3, 3, 4797723, NEWID(), 14.250),
(410, 2, 2, 4797723, NEWID(), 4.780),
(440, 2, 2, 4797723, NEWID(), 21.000),
(480, 1, 1, 4797723, NEWID(), 3.500),
(1290, 2, 2, 4797723, NEWID(), 39.280),
(470, 1, 1, 4797723, NEWID(), 8.500),
(280, 3, 3, 4797723, NEWID(), 16.500),
(10, 2, 2, 4797723, NEWID(), 10.700),
(500, 2, 2, 4797723, NEWID(), 6.600),
(290, 1, 1, 4797713, NEWID(), 0.540),
(40, 2, 2, 4797713, NEWID(), 33.800)

SELECT 
    [dispatchPallet].[SHIPMENT_ID], 
    SUM([pickLog].[QTY_SUF]) AS KOLLI,
    COUNT(DISTINCT [pickLog].[LINE_NUM]) AS LINES,
    SUM([pickLog].[WEIGHT_GROSS]) AS PICKED_WEIGHT, 
    COUNT(DISTINCT [pickLog].[PICK_PALL_NUM]) AS PALLETS,
    COUNT(DISTINCT CASE WHEN [dispatchPallet].[PALLET_PLACEMENT] = 'B' THEN [dispatchPallet].[PICK_PALL_NUM] ELSE NULL END) AS BOTTOM_PALLETS       
FROM 
    @dispatchPallet dispatchPallet 
    INNER JOIN @pickLog pickLog ON [dispatchPallet].[PICK_PALL_NUM] = [pickLog].[PICK_PALL_NUM]
GROUP BY 
    [dispatchPallet].[SHIPMENT_ID]

-- Expected output:
-- SHIPMENT_ID  KOLLI   LINES   PICKED_WEIGHT   PALLETS BOTTOM_PALLETS
-- SHIPMENT-1   54.00   1       1070.280        1       1
-- SHIPMENT-2   24.00   1       471.360         1       1
-- SHIPMENT-3   12.00   1       237.960         1       1
-- SHIPMENT-4   25.00   1       102.750         1       1
-- SHIPMENT-5   3.00    1       40.650          1       1
-- SHIPMENT-6   16.00   2       316.640         1       1
-- SHIPMENT-7   60.00   1       330.000         2       2
-- SHIPMENT-8   18.00   9       125.110         1       1
-- SHIPMENT-9   3.00    2       34.340          1       1

Answer 1

您至少应该在

上创建主要约束

ALTER TABLE @dispatchPallet TABLE   ADD PRIMARY KEY (PICK_PALL_NUM);

外键约束为

ALTER TABLE @pickLog TABLE  ADD foreign key (PICK_PALL_NUM) references @dispatchPallet(PICK_PALL_NUM)

还在

上创建一个唯一索引

CREATE UNIQUE NONCLUSTERED INDEX idx_PALLET_PLACEMENT_notnull
ON @dispatchPallet(PALLET_PLACEMENT)
WHERE PALLET_PLACEMENT IS NOT NULL;

Answer 2

您的查询很简单，并且没有太多优化空间。您应该检查您至少dispatchPallet SHIPMENT_ID和pickLog PICK_PALL_NUM上的索引。这些将是您查询的最佳选择：

CREATE NONCLUSTERED INDEX NCI_dispatchPallet_shipment_ID 
    ON dispatchPallet (SHIPMENT_ID, PICK_PALL_NUM)
    INCLUDE (PALLET_PLACEMENT)


CREATE NONCLUSTERED INDEX NCI_pickLog_pick_pall_num 
    ON pickLog (PICK_PALL_NUM)
    INCLUDE (QTY_SUF, LINE_NUM, WEIGHT_GROSS)

您还应该验证您是否需要COUNT DISTINCT（不同的是昂贵的操作）。

最后但并非最不重要的是，您应该检查访问视图的方式;如果您正在过滤它，加入它等等。这些其他条件可能会生成不同的查询计划，如果管理不正确，您的性能会下降（即使使用正确的索引！）。

Answer 3

对于初学者来说，这些表上应该有主键和外键，以便此查询可以进行索引搜索/扫描（上面的paparazzo评论），而不是全表搜索/扫描。

除了bigint / int之外，uniqueidentifier的目的是什么？

优化汇总查询

3 个答案: