按顺序消除分区
Ste*_*old 4 sql-server partitioning azure-sql-database
我有一个分区表和分区消除问题。
设置
我的表很简单，看起来像这样：
CREATE TABLE [data].[OperationData](
[DataLoggerID] [bigint] NOT NULL,
[End] [datetime2](7) NOT NULL
CONSTRAINT [PK_OperationData] PRIMARY KEY CLUSTERED
(
[DataLoggerID] ASC,
[End] ASC
))
该表[End]按日期按列进行分区。
问题
我执行一个简单的查询以逐[End]列查找最新的行。
SELECT TOP(1)
[OperationData].[DataLoggerID],
[OperationData].[End]
FROM [data].[OperationData]
WHERE [OperationData].[DataLoggerID] = 30217
ORDER BY [OperationData].[End] DESC
问题是执行计划显示所有分区都被搜索过了，没有发生分区消除。
题
此查询是否可以消除分区？当是的时候为什么不使用它？
多一点...
我无法将计划粘贴到“粘贴计划”上（显示错误），因此我将其复制到此处。
<?xml version="1.0" encoding="utf-16"?>
<ShowPlanXML xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xsd="http://www.w3.org/2001/XMLSchema" Version="1.520" Build="15.0.400.390" xmlns="http://schemas.microsoft.com/sqlserver/2004/07/showplan">
<BatchSequence>
<Batch>
<Statements>
<StmtSimple StatementCompId="1" StatementEstRows="1" StatementId="1" StatementOptmLevel="FULL" CardinalityEstimationModelVersion="130" StatementSubTreeCost="26.2112" StatementText="SELECT TOP(1)
 [OperationData].[DataLoggerID],
 [OperationData].[End]
FROM [data].[OperationData] AS [OperationData] 
WHERE [OperationData].[DataLoggerID] = 30217 
ORDER BY [OperationData].[End] DESC" StatementType="SELECT" QueryHash="0xCA9C86DFB3F9CA37" QueryPlanHash="0xB953A3686656A27D" RetrievedFromCache="true" StatementSqlHandle="0x090060DDDE8D2D2A050D6AF14FDF6902EDD10000000000000000000000000000000000000000000000000000" DatabaseContextSettingsId="7" ParentObjectId="0" StatementParameterizationType="0" SecurityPolicyApplied="false">
<StatementSetOptions ANSI_NULLS="true" ANSI_PADDING="true" ANSI_WARNINGS="true" ARITHABORT="true" CONCAT_NULL_YIELDS_NULL="true" NUMERIC_ROUNDABORT="false" QUOTED_IDENTIFIER="true" />
<QueryPlan DegreeOfParallelism="0" NonParallelPlanReason="MaxDOPSetToOne" MemoryGrant="1024" CachedPlanSize="280" CompileTime="6783" CompileCPU="52" CompileMemory="7848">
<MissingIndexes>
<MissingIndexGroup Impact="95.7193">
<MissingIndex Database="[powerdoo]" Schema="[data]" Table="[OperationData]">
<ColumnGroup Usage="EQUALITY">
<Column Name="[DataLoggerID]" ColumnId="1" />
</ColumnGroup>
</MissingIndex>
</MissingIndexGroup>
</MissingIndexes>
<Warnings>
<MemoryGrantWarning GrantWarningKind="Excessive Grant" RequestedMemory="1024" GrantedMemory="1024" MaxUsedMemory="16" />
</Warnings>
<MemoryGrantInfo SerialRequiredMemory="16" SerialDesiredMemory="24" RequiredMemory="16" DesiredMemory="24" RequestedMemory="1024" GrantWaitTime="0" GrantedMemory="1024" MaxUsedMemory="16" MaxQueryMemory="3104936" />
<OptimizerHardwareDependentProperties EstimatedAvailableMemoryGrant="5872010" EstimatedPagesCached="1101001" EstimatedAvailableDegreeOfParallelism="1" MaxCompileMemory="9384056" />
<WaitStats>
<Wait WaitType="SOS_SCHEDULER_YIELD" WaitTimeMs="805" WaitCount="19" />
<Wait WaitType="RESOURCE_GOVERNOR_IDLE" WaitTimeMs="15" WaitCount="8" />
</WaitStats>
<QueryTimeStats CpuTime="76" ElapsedTime="846" />
<RelOp AvgRowSize="23" EstimateCPU="1.00318" EstimateIO="0.0112613" EstimateRebinds="0" EstimateRewinds="0" EstimatedExecutionMode="Row" EstimateRows="1" LogicalOp="TopN Sort" NodeId="0" Parallel="false" PhysicalOp="Sort" EstimatedTotalSubtreeCost="26.2112">
<OutputList>
<ColumnReference Database="[powerdoo]" Schema="[data]" Table="[OperationData]" Alias="[OperationData]" Column="DataLoggerID" />
<ColumnReference Database="[powerdoo]" Schema="[data]" Table="[OperationData]" Alias="[OperationData]" Column="End" />
</OutputList>
<MemoryFractions Input="1" Output="1" />
<RunTimeInformation>
<RunTimeCountersPerThread Thread="0" ActualRebinds="1" ActualRewinds="0" ActualRows="1" Batches="0" ActualEndOfScans="1" ActualExecutions="1" ActualExecutionMode="Row" ActualElapsedms="845" ActualCPUms="76" ActualScans="0" ActualLogicalReads="0" ActualPhysicalReads="0" ActualReadAheads="0" ActualLobLogicalReads="0" ActualLobPhysicalReads="0" ActualLobReadAheads="0" InputMemoryGrant="1024" OutputMemoryGrant="1024" UsedMemoryGrant="16" />
</RunTimeInformation>
<TopSort Distinct="false" Rows="1">
<OrderBy>
<OrderByColumn Ascending="false">
<ColumnReference Database="[powerdoo]" Schema="[data]" Table="[OperationData]" Alias="[OperationData]" Column="End" />
</OrderByColumn>
</OrderBy>
<RelOp AvgRowSize="23" EstimateCPU="1.22175" EstimateIO="23.975" EstimateRebinds="0" EstimateRewinds="0" EstimatedExecutionMode="Row" EstimateRows="15680.7" EstimatedRowsRead="15680.7" LogicalOp="Clustered Index Seek" NodeId="1" Parallel="false" Partitioned="true" PhysicalOp="Clustered Index Seek" EstimatedTotalSubtreeCost="25.1968" TableCardinality="81308300">
<OutputList>
<ColumnReference Database="[powerdoo]" Schema="[data]" Table="[OperationData]" Alias="[OperationData]" Column="DataLoggerID" />
<ColumnReference Database="[powerdoo]" Schema="[data]" Table="[OperationData]" Alias="[OperationData]" Column="End" />
</OutputList>
<RunTimeInformation>
<RunTimeCountersPerThread Thread="0" ActualRows="49050" ActualRowsRead="49050" Batches="0" ActualEndOfScans="1" ActualExecutions="1" ActualExecutionMode="Row" ActualElapsedms="730" ActualCPUms="64" ActualScans="7672" ActualLogicalReads="2533" ActualPhysicalReads="0" ActualReadAheads="65" ActualLobLogicalReads="0" ActualLobPhysicalReads="0" ActualLobReadAheads="0" />
</RunTimeInformation>
<RunTimePartitionSummary>
<PartitionsAccessed PartitionCount="7672">
<PartitionRange Start="1" End="7672" />
</PartitionsAccessed>
</RunTimePartitionSummary>
<IndexScan Ordered="true" ScanDirection="FORWARD" ForcedIndex="false" ForceSeek="false" ForceScan="false" NoExpandHint="false" Storage="RowStore">
<DefinedValues>
<DefinedValue>
<ColumnReference Database="[powerdoo]" Schema="[data]" Table="[OperationData]" Alias="[OperationData]" Column="DataLoggerID" />
</DefinedValue>
<DefinedValue>
<ColumnReference Database="[powerdoo]" Schema="[data]" Table="[OperationData]" Alias="[OperationData]" Column="End" />
</DefinedValue>
</DefinedValues>
<Object Database="[powerdoo]" Schema="[data]" Table="[OperationData]" Index="[PK_OperationData]" Alias="[OperationData]" IndexKind="Clustered" Storage="RowStore" />
<SeekPredicates>
<SeekPredicateNew>
<SeekKeys>
<StartRange ScanType="GE">
<RangeColumns>
<ColumnReference Column="PtnId1000" />
</RangeColumns>
<RangeExpressions>
<ScalarOperator ScalarString="(1)">
<Const ConstValue="(1)" />
</ScalarOperator>
</RangeExpressions>
</StartRange>
<EndRange ScanType="LE">
<RangeColumns>
<ColumnReference Column="PtnId1000" />
</RangeColumns>
<RangeExpressions>
<ScalarOperator ScalarString="(7672)">
<Const ConstValue="(7672)" />
</ScalarOperator>
</RangeExpressions>
</EndRange>
</SeekKeys>
<SeekKeys>
<Prefix ScanType="EQ">
<RangeColumns>
<ColumnReference Database="[powerdoo]" Schema="[data]" Table="[OperationData]" Alias="[OperationData]" Column="DataLoggerID" />
</RangeColumns>
<RangeExpressions>
<ScalarOperator ScalarString="(30217)">
<Const ConstValue="(30217)" />
</ScalarOperator>
</RangeExpressions>
</Prefix>
</SeekKeys>
</SeekPredicateNew>
</SeekPredicates>
</IndexScan>
</RelOp>
</TopSort>
</RelOp>
</QueryPlan>
</StmtSimple>
</Statements>
</Batch>
</BatchSequence>
</ShowPlanXML>
Pau*_*ite 11
优化器可以为此查询生成最佳计划：
示例架构
CREATE PARTITION FUNCTION PF (datetime2(7)) AS
RANGE RIGHT FOR VALUES
(
'20180101','20180102','20180103','20180104','20180105','20180106','20180107','20180108','20180109','20180110',
'20180111','20180112','20180113','20180114','20180115','20180116','20180117','20180118','20180119','20180120',
'20180121','20180122','20180123','20180124','20180125','20180126','20180127','20180128','20180129','20180130',
'20180131','20180201','20180202','20180203','20180204','20180205','20180206','20180207','20180208','20180209',
'20180210','20180211','20180212','20180213','20180214','20180215','20180216','20180217','20180218','20180219',
'20180220','20180221','20180222','20180223','20180224','20180225','20180226','20180227','20180228','20180301',
'20180302','20180303','20180304','20180305','20180306','20180307','20180308','20180309','20180310','20180311',
'20180312','20180313','20180314','20180315','20180316','20180317','20180318','20180319','20180320','20180321',
'20180322','20180323','20180324','20180325','20180326','20180327','20180328','20180329','20180330','20180331',
'20180401','20180402','20180403','20180404','20180405','20180406','20180407','20180408','20180409','20180410',
'20180411','20180412','20180413','20180414','20180415','20180416','20180417','20180418','20180419','20180420',
'20180421','20180422','20180423','20180424','20180425','20180426','20180427','20180428','20180429','20180430',
'20180501','20180502','20180503','20180504','20180505','20180506','20180507','20180508','20180509','20180510',
'20180511','20180512','20180513','20180514','20180515','20180516','20180517','20180518','20180519','20180520',
'20180521','20180522','20180523','20180524','20180525','20180526','20180527','20180528','20180529','20180530',
'20180531','20180601','20180602','20180603','20180604','20180605','20180606','20180607','20180608','20180609',
'20180610','20180611','20180612','20180613','20180614','20180615','20180616','20180617','20180618','20180619',
'20180620','20180621','20180622','20180623','20180624','20180625','20180626','20180627','20180628','20180629',
'20180630','20180701'
);
CREATE PARTITION SCHEME PS
AS PARTITION PF
ALL TO ([PRIMARY]);
样本数据
CREATE TABLE [dbo].[OperationData]
(
[DataLoggerID] [bigint] NOT NULL,
[End] [datetime2](7) NOT NULL
CONSTRAINT [PK_OperationData] PRIMARY KEY CLUSTERED
(
[DataLoggerID] ASC,
[End] ASC
) ON PS ([End])
);
-- Add one million rows
WITH
L0 AS (SELECT 1 AS c UNION ALL SELECT 1),
L1 AS (SELECT 1 AS c FROM L0 AS A CROSS JOIN L0 AS B),
L2 AS (SELECT 1 AS c FROM L1 AS A CROSS JOIN L1 AS B),
L3 AS (SELECT 1 AS c FROM L2 AS A CROSS JOIN L2 AS B),
L4 AS (SELECT 1 AS c FROM L3 AS A CROSS JOIN L3 AS B),
L5 AS (SELECT 1 AS c FROM L4 AS A CROSS JOIN L4 AS B),
Nums AS (SELECT ROW_NUMBER() OVER(ORDER BY (SELECT NULL)) AS n FROM L5)
INSERT dbo.OperationData
WITH (TABLOCKX)
(DataLoggerID, [End])
SELECT
Nums.n % 50000,
DATEADD(DAY, Nums.n % 366, '20180101')
FROM Nums
WHERE
Nums.n <= 1000 * 1000;
原始查询
SELECT TOP (1)
[OperationData].[DataLoggerID],
[OperationData].[End]
FROM [dbo].[OperationData]
WHERE [OperationData].[DataLoggerID] = 30217
ORDER BY [OperationData].[End] DESC;
执行计划显示每个分区一个向后排序的搜索，并返回找到的第一个符合条件的行：
这正是您所希望的：从最高数字向下处理每个分区，立即寻找给定的最高End日期和时间DataLoggerID，并在找到第一行后停止。
该计划是在 Microsoft SQL Azure 上找到的：
Version="1.520" Build="15.0.400.390"
我还验证了优化器为从 SQL Server 2008 R2 到 SQL Server 2017 的所有版本生成相同的最佳计划。这包括对 15,000 个分区（如果可用）的测试。
解决方法
否则，您可以更明确地编写查询逻辑：
CREATE TABLE #Partitions (partition_number integer PRIMARY KEY);
INSERT #Partitions
(partition_number)
SELECT
P.partition_number
FROM sys.partitions AS P
WHERE
P.[object_id] = OBJECT_ID(N'dbo.OperationData', 'U')
SELECT TOP (1)
LocalMinimum.DataLoggerID,
LocalMinimum.[End]
FROM #Partitions AS P
CROSS APPLY
(
SELECT TOP (1)
OD.DataLoggerID,
OD.[End]
FROM dbo.OperationData AS OD
WHERE
$PARTITION.PF(OD.[End]) = P.partition_number
AND OD.DataLoggerID = 30217
ORDER BY
OD.[End] DESC
) AS LocalMinimum
WHERE
LocalMinimum.DataLoggerID = 30217
ORDER BY
P.partition_number DESC,
LocalMinimum.[End] DESC;
这给出了将分区号加载到临时表中的以下计划：
这个针对每个分区查询的计划：
再次以相反的顺序处理分区，直到找到第一行。