SQL Server 2012中快速插入批量数据的示例及疑惑

今天在做一个案例演示时，在SQL Server 2012中使用Insert语句插入1万条数据，结果遇到了一个奇怪的现象，现将过程分享出来，以供有兴趣的同学参考。

附：我的测试环境为：
SQL Server 2012，命名实例

Microsoft SQL Server 2012 - 11.0.2100.60 (Intel X86) 
Feb 10 2012 19:13:17 
Copyright (c) Microsoft Corporation
Enterprise Edition on Windows NT 6.0 <X86> (Build 6002: Service Pack 2)

创建示例数据库

IF OBJECT_ID('DemoPager2012') IS NOT NULL
DROP DataBase DemoPager2012
GO

CREATE Database DemoPager2012
GO

USE DemoPager2012
GO

示例表，该表只有四个字段。

/*
Setup script to create the sample table and fill it with
sample data.
*/
IF OBJECT_ID('Customers','U') IS NOT NULL
DROP TABLE Customers

CREATE TABLE Customers ( CustomerID INT primary key identity(1,1),
CustomerNumber CHAR(4),
CustomerName VARCHAR(50),
CustomerCity VARCHAR(20) )
GO

现在展示批量插入10000条数据到该表中，语句如下：

TRUNCATE table Customers
GO

----清除干扰查询
DBCC DROPCLEANBUFFERS
DBCC FREEPROCCACHE

SET STATISTICS IO ON;
SET STATISTICS TIME ON;
GO

DECLARE @d Datetime
SET @d=getdate();

declare @i int=1
while @i<=10000
begin
INSERT INTO Customers (CustomerNumber, CustomerName,
CustomerCity)
SELECT REPLACE(STR(@i, 4), ' ', '0'),'Customer ' + STR(@i,6),
CHAR(65 + (@i % 26)) + '-City'
set @i=@i+1
end

select [语句执行花费时间(毫秒)]=datediff(ms,@d,getdate())

SET STATISTICS IO OFF ;
SET STATISTICS TIME OFF;
GO

该插入语句在SQL Server 2008 r2版本和SQL Server 2012版本中，测试结果如下：

令我惊讶的是，SQL Server 2012居然耗时达到5分多钟，而SQL Server 2008R2版，只需要大约6秒钟。更令人费解的是：查询的I/O统计和elapsed time，在这两个版本中几乎一样。对此异象，我只能理解为每次Insert时的毫秒级精度可能不足以度量该次操作带来的细小差距，然而累积起来就非常可观了。

解决方案一：使用 Set NoCount On，效果立竿见影

TRUNCATE table Customers
GO

----清除干扰查询
DBCC DROPCLEANBUFFERS
DBCC FREEPROCCACHE

SET STATISTICS IO ON;
SET STATISTICS TIME ON;
GO

DECLARE @d Datetime
SET @d=getdate();
set nocount on
declare @i int=1
while @i<=10000
begin
INSERT INTO Customers (CustomerNumber, CustomerName,
CustomerCity)
SELECT REPLACE(STR(@i, 4), ' ', '0'),'Customer ' + STR(@i,6),
CHAR(65 + (@i % 26)) + '-City'
set @i=@i+1
end

select [语句执行花费时间(毫秒)]=datediff(ms,@d,getdate())

SET STATISTICS IO OFF ;
SET STATISTICS TIME OFF;
GO

Set NoCount On（http://msdn.microsoft.com/zh-cn/library/ms189837.aspx）的作用：使返回的结果中不包含有关受 Transact-SQL 语句影响的行数的信息。这在批量插入时将显著提高性能。至于 本例中，为什么SQL Server 2008 R2版中却不受该开关影响，希望知道的同学不吝赐教，非常感谢。

改进解决方案二：使用 Set NoCount On+Transaction

TRUNCATE table Customers
GO

----清除干扰查询
DBCC DROPCLEANBUFFERS
DBCC FREEPROCCACHE

 

SET STATISTICS IO ON;
SET STATISTICS TIME ON;
GO

DECLARE @d Datetime
SET @d=getdate();
set nocount on
declare @i int=1
BEGIN TRANSACTION
while @i<=10000
begin
INSERT INTO Customers (CustomerNumber, CustomerName,
CustomerCity)
SELECT REPLACE(STR(@i, 4), ' ', '0'),'Customer ' + STR(@i,6),
CHAR(65 + (@i % 26)) + '-City'
set @i=@i+1
end
COMMIT
select [语句执行花费时间(毫秒)]=datediff(ms,@d,getdate())

SET STATISTICS IO OFF ;
SET STATISTICS TIME OFF;
GO

解决方案三：使用递归CTE插入

TRUNCATE table Customers
GO

DBCC DROPCLEANBUFFERS
DBCC FREEPROCCACHE

SET STATISTICS IO ON;
SET STATISTICS TIME ON;
GO

DECLARE @d Datetime
SET @d=getdate();

/*****运用CTE递归插入，速度较快，邀月注***********************/
WITH Seq (num,CustomerNumber, CustomerName, CustomerCity) AS
(SELECT 1,cast('0000'as CHAR(4)),cast('Customer 0' AS NVARCHAR(50)),cast('X-City' as NVARCHAR(20))
UNION ALL
SELECT num + 1,Cast(REPLACE(STR(num, 4), ' ', '0') AS CHAR(4)),
cast('Customer ' + STR(num,6) AS NVARCHAR(50)),
cast(CHAR(65 + (num % 26)) + '-City' AS NVARCHAR(20))
FROM Seq
WHERE num <= 10000
)
INSERT INTO Customers (CustomerNumber, CustomerName, CustomerCity)
SELECT CustomerNumber, CustomerName, CustomerCity
FROM Seq
OPTION (MAXRECURSION 0)

select [语句执行花费时间(毫秒)]=datediff(ms,@d,getdate())

SET STATISTICS IO OFF ;
SET STATISTICS TIME OFF;
GO

小结：SQL Server 2012中批量插入数据时，请记得Set NoCount ON，并尽可能加上Transaction，当然，推荐使用CTE，这可能会带来性能上的巨大提升。

 邀月补充：

后来与微软亚太工程师多次沟通，得出初步结论：

在不打开“set nocount on”时，SSMS 2012与SSMS 2008r2版本的UI在执行效率上可能有极大差异，而与SQL Server引擎没有明显相关。

转自邀月工作室