SQLServer 2000 升级到 SQLServer 2008 性能之需要注意的地方之
今天在 相同环境测试 2000 和 2008 性能 让我意外的是 2008 明显比2000 慢很多,因为不能简单的升级,sql语句也需要优化
测试sql:代码如下:
SET STATISTICS IO ON
SET STATISTICS TIME ON
SELECT COUNT(1)
FROM dbo.tbtext a
INNER LOOP JOIN dbo.tbtext b
ON a.id = b.id option (maxdop 1)
SET STATISTICS IO Off
SET STATISTICS TIME Off
表结构:
代码如下:
CREATE TABLE [dbo].[tbtext](
[id] [int] IDENTITY(1,1) NOT NULL,
[VALUE] [int] NULL
) ON [PRIMARY]
单这句测试,看执行计划根本看不出区别。
|--Compute Scalar(DEFINE:([Expr1006]=CONVERT_IMPLICIT(int,[Expr1009],0)))
|--Stream Aggregate(DEFINE:([Expr1009]=Count(*)))
|--Nested Loops(Inner Join, WHERE:([northwind].[dbo].[tbtext].[id] as [b].[id]=[northwind].[dbo].[tbtext].[id] as [a].[id]))
|--Table Scan(OBJECT:([northwind].[dbo].[tbtext] AS [a]))
|--Table Spool
|--Table Scan(OBJECT:([northwind].[dbo].[tbtext] AS [b]))
2008r2:
代码如下:
/*
警告: 由于使用了本地联接提示,联接次序得以强制实施。
表 'tbtext'。扫描计数 1,逻辑读取 46 次
(1 行受影响)
表 'Worktable'。扫描计数 1,逻辑读取 290098 次,物理读取 0 次,预读 0 次,lob 逻辑读取 0 次,lob 物理读取 0 次,lob 预读 0 次。
表 'tbtext'。扫描计数 2,逻辑读取 262 次,物理读取 0 次,预读 0 次,lob 逻辑读取 0 次,lob 物理读取 0 次,lob 预读 0 次。
(1 行受影响)
SQL Server 执行时间:
CPU 时间 = 32828 毫秒,占用时间 = 32846 毫秒。
SQL Server 执行时间:
CPU 时间 = 0 毫秒,占用时间 = 0 毫秒。
*/
2000sp4:
代码如下:
/*
警告: 由于使用了局部联接提示,所以联接次序得以强制实施。
表 'tbtext'。扫描计数 1,逻辑读 131 次,物理读 0 次,预读 0 次。
SQL Server 执行时间:
CPU 时间 = 0 毫秒,耗费时间 = 0 毫秒。
表 'Worktable'。扫描计数 9999,逻辑读 180001 次,物理读 0 次,预读 0 次。
表 'tbtext'。扫描计数 2,逻辑读 262 次,物理读 0 次,预读 138 次。
SQL Server 执行时间:
CPU 时间 = 17188 毫秒,耗费时间 = 17261 毫秒。
(1 行受影响)
SQL Server 执行时间:
CPU 时间 = 0 毫秒,耗费时间 = 0 毫秒。
*/
比较2000 和 2008的执行就能发现 2008 的cpu 时间明显比 2000 高,2008 的worktable 逻辑读取量,比2000的高,
这个有个worktable 的扫描技术,2000的是9999,2008的是1,这个让人难免有的疑惑是什么情况,都是nest loop,worktable 扫描不应该是1才对。
性能差怎么大会不会是 worktable 搞的鬼呢?
那么就开始调节,过滤id 会有啥发现呢?
代码如下:
SET STATISTICS IO ON
SET STATISTICS TIME ON
SELECT COUNT(1)
FROM dbo.tbtext a
INNER LOOP JOIN dbo.tbtext b
ON a.id = b.id
WHERE a.id SELECT COUNT(1)
FROM dbo.tbtext a
SET STATISTICS IO Off
SET STATISTICS TIME Off
2008r2:
SELECT COUNT(1) FROM dbo.tbtext a INNER LOOP JOIN dbo.tbtext b ON a.id = b.id WHERE a.id |--Compute Scalar(DEFINE:([Expr1006]=CONVERT_IMPLICIT(int,[Expr1009],0)))
|--Stream Aggregate(DEFINE:([Expr1009]=Count(*)))
|--Nested Loops(Inner Join, WHERE:([northwind].[dbo].[tbtext].[id] as [b].[id]=[northwind].[dbo].[tbtext].[id] as [a].[id]))
|--Table Scan(OBJECT:([northwind].[dbo].[tbtext] AS [a]), WHERE:([northwind].[dbo].[tbtext].[id] as [a].[id]|--Table Spool
|--Table Scan(OBJECT:([northwind].[dbo].[tbtext] AS [b]), WHERE:([northwind].[dbo].[tbtext].[id] as [b].[id] 代码如下:
表 'Worktable'。扫描计数 1,逻辑读取 6006 次,物理读取 0 次,预读 0 次,lob 逻辑读取 0 次,lob 物理读取 0 次,lob 预读 0 次。
表 'tbtext'。扫描计数 2,逻辑读取 262 次,物理读取 0 次,预读 0 次,lob 逻辑读取 0 次,lob 物理读取 0 次,lob 预读 0 次。
2000sp4:
|--Compute Scalar(DEFINE:([Expr1002]=Convert([Expr1006])))
|--Stream Aggregate(DEFINE:([Expr1006]=Count(*)))
|--Nested Loops(Inner Join, WHERE:([b].[id]=[a].[id]))
|--Table Scan(OBJECT:([Northwind].[dbo].[tbtext] AS [a]), WHERE:([a].[id]|--Table Spool
|--Table Scan(OBJECT:([Northwind].[dbo].[tbtext] AS [b]))
代码如下:
表 'Worktable'。扫描计数 999,逻辑读 27001 次,物理读 0 次,预读 0 次。
表 'tbtext'。扫描计数 2,逻辑读 262 次,物理读 0 次,预读 0次。
进入 lazy spool的数据完全不一样了,2008 只是进入了1000 条数据,但是2000 全都进去了。
在逻辑读上面 2008 明显低于 2000. cpu时间也明显比2000少。
通过调节id 的值,2000 我推出了一个公式 逻辑读= 10001+(17*n) ,
但是2008的算法十分奇怪,
当n 当 3862000的逻辑读是线性增长,2008 是分段的线性增长,每个分段 f '(x) 都不一样。
2008 的lazy spool适合选择度高的,选择度低的时候完全不行。
从2000到2008 不单单是多了sqlos和表面上的一些功能,很多执行计划的操作符都被重写了,像lazy spool 。
所以在升级到2008 之前,
各位朋友,是否都应该重写一下sql 呢?单单优化 索引 已经解决不了根本问题了。

Hot AI Tools

Undresser.AI Undress
AI-powered app for creating realistic nude photos

AI Clothes Remover
Online AI tool for removing clothes from photos.

Undress AI Tool
Undress images for free

Clothoff.io
AI clothes remover

AI Hentai Generator
Generate AI Hentai for free.

Hot Article

Hot Tools

Notepad++7.3.1
Easy-to-use and free code editor

SublimeText3 Chinese version
Chinese version, very easy to use

Zend Studio 13.0.1
Powerful PHP integrated development environment

Dreamweaver CS6
Visual web development tools

SublimeText3 Mac version
God-level code editing software (SublimeText3)

Hot Topics



Both vivox100s and x100 mobile phones are representative models in vivo's mobile phone product line. They respectively represent vivo's high-end technology level in different time periods. Therefore, the two mobile phones have certain differences in design, performance and functions. This article will conduct a detailed comparison between these two mobile phones in terms of performance comparison and function analysis to help consumers better choose the mobile phone that suits them. First, let’s look at the performance comparison between vivox100s and x100. vivox100s is equipped with the latest

In this tutorial, we will help you reveal hidden performance overlays in Windows 11. Using Windows 11's Performance Overlay feature, you'll be able to monitor your system resources in real time. You can view real-time CPU usage, disk usage, GPU usage, RAM usage, etc. on your computer screen. This is convenient when you are playing games or using large graphics programs (such as video editors) and need to check how much system performance is affected when using a specific program. While there are some excellent free software available for monitoring system performance, and some built-in tools like Resource Monitor can be used to check system performance, the performance overlay feature also has its advantages. For example, you don't need to leave the program or app you're currently using or

Windows 10 vs. Windows 11 performance comparison: Which one is better? With the continuous development and advancement of technology, operating systems are constantly updated and upgraded. As one of the world's largest operating system developers, Microsoft's Windows series of operating systems have always attracted much attention from users. In 2021, Microsoft released the Windows 11 operating system, which triggered widespread discussion and attention. So, what is the difference in performance between Windows 10 and Windows 11? Which

The Windows operating system has always been one of the most widely used operating systems on personal computers, and Windows 10 has long been Microsoft's flagship operating system until recently when Microsoft launched the new Windows 11 system. With the launch of Windows 11 system, people have become interested in the performance differences between Windows 10 and Windows 11 systems. Which one is better between the two? First, let’s take a look at W

In the era of mobile Internet, smartphones have become an indispensable part of people's daily lives. The performance of smartphones often directly determines the quality of user experience. As the "brain" of a smartphone, the performance of the processor is particularly important. In the market, the Qualcomm Snapdragon series has always been a representative of strong performance, stability and reliability, and recently Huawei has also launched its own Kirin 8000 processor, which is said to have excellent performance. For ordinary users, how to choose a mobile phone with strong performance has become a key issue. Today we will

PHP and Go are two commonly used programming languages, and they have different characteristics and advantages. Among them, performance difference is an issue that everyone is generally concerned about. This article will compare PHP and Go languages from a performance perspective, and demonstrate their performance differences through specific code examples. First, let us briefly introduce the basic features of PHP and Go language. PHP is a scripting language originally designed for web development. It is easy to learn and use and is widely used in the field of web development. The Go language is a compiled language developed by Google.

Ollama is a super practical tool that allows you to easily run open source models such as Llama2, Mistral, and Gemma locally. In this article, I will introduce how to use Ollama to vectorize text. If you have not installed Ollama locally, you can read this article. In this article we will use the nomic-embed-text[2] model. It is a text encoder that outperforms OpenAI text-embedding-ada-002 and text-embedding-3-small on short context and long context tasks. Start the nomic-embed-text service when you have successfully installed o

Performance comparison of different Java frameworks: REST API request processing: Vert.x is the best, with a request rate of 2 times SpringBoot and 3 times Dropwizard. Database query: SpringBoot's HibernateORM is better than Vert.x and Dropwizard's ORM. Caching operations: Vert.x's Hazelcast client is superior to SpringBoot and Dropwizard's caching mechanisms. Suitable framework: Choose according to application requirements. Vert.x is suitable for high-performance web services, SpringBoot is suitable for data-intensive applications, and Dropwizard is suitable for microservice architecture.
