[转帖]UsingDiskspdforSQLServer --- 改天做下练习 官方文档.
Using DiskSpd in SQL Server environments Writer: Robert Beene Contributors: Jose Barreto, Ramu Konidena Technical Reviewers: Robert Dorr, Mike Zawadzki, Nitin Verma, Dan Lovinger Applies To: All Versions of SQL Server except Microsoft Azure SQL Database (MASD) Topics
|
||||||||||||||||||||||||
BACKGROUND |
A number of tools are available to stress and validate the functionality of I/O subsystems. Some of these tools actually simulate SQL Server I/O patterns, and others let you control the I/O patterns submitted to the subsystem. There are tools that confirm the functional validity of a configuration, while others are used only to test performance capacity. Software that performs a validation test simulates a read and write pattern, but it also verifies that the information was correctly read and written. This type of software is often used to help find root causes of corruption or to help determine whether drivers and hardware are functioning as expected when they’re performing I/O. There is likely to be a speed component to running this test, but this test is more about accuracy than speed. For the purposes of this document, a performance test tool focuses on simulating the READ and WRITE pattern, but it focuses even more heavily on the throughput and speed at which those operations take place (Load Generator). A performance test determines whether the hardware is capable of meeting the parameters at which it was designed to handle I/O throughput. The following table compares the most common tools that are currently used to test I/O subsystems in the context of SQL Server. Table 1 Tools used to test I/O subsystems
SQLIOSim For the |
|||||||||||||||||||||||
INFORMATION |
DiskSpd.exe You can use DiskSpd |
|||||||||||||||||||||||
FEATURES |
Some of the
|
|||||||||||||||||||||||
Open Source DiskSpd is
|
||||||||||||||||||||||||
Download details DISKSPD is To download DiskSpd, |
||||||||||||||||||||||||
CONTENTS |
The DiskSpd download Figure 1 |
|||||||||||||||||||||||
SYNTAX |
To display a DiskSpd DiskSpd [options] target1 [ target2 [ The following DiskSpd –d300 Test targets All available options and parameters are enumerated in |
|||||||||||||||||||||||
WARNINGS & PRECAUTIONS |
WARNING: When WARNING: If Note: Be aware that uppercase and lowercase Note: Make sure there’s nothing else Note: Run DiskSpd from an elevated Note: You can cancel a DiskSpd run at any Note: The iB notation is an international • 1KiB = 210 = 1,024 bytes • 1MiB = 1024 KiB = 220 = • 1GiB = 1024 MiB = 230 = |
|||||||||||||||||||||||
USAGE |
For complete In order to The following Note: Make sure that you always use a
Now that we The following DiskSpd.exe -c1000G –d300 -r -w0 -t8 -o8
Note: The test file that’s created for
EXAMPLES Note: All A single target file sized at 1000 GiB 8 threads for data files 4 threads for Log files (1 per NUMA node 300 Second duration 32 outstanding IOs for data files, 8 up Your test may require modifying these Note: |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
ANALYSIS |
For complete By default, Before For example, assume Figure 2
Understanding It’s For example,
There is a penalty MAX_WRITE * NUMBER_OF_DRIVES / The following
Even though each drive can handle up to 450 MB/sec
Now that we have an understanding of the possible max Repeats Command Line: Note: For this test, –L for latency Lists Input timespan: 1 ------------- duration: 10s warm up time: 5s cool down time: 0s measuring latency random seed: 0 path: 'F:\testfile.dat' think time: 0ms burst size: 0 software and hardware write performing read test block size: 8192 using random I/O (alignment: number of outstanding I/O thread stride size: 0 threads per file: 8 using I/O Completion Ports IO priority: normal Results for ******************************************************************************* Shows actual test thread count: 8 proc count: 12 Shows CPU | Usage | ------------------------------------------- 0| 1| 2| 3| 4| 5| 6| 7| 8| 9| 10| 11| ------------------------------------------- avg.| 2.43%| Gives Total IO thread | bytes | ----------------------------------------------------------------------------------------------------- 0 | 5341184 | 652 | 0.51 | 65.20 | 1 | 5865472 | 716 | 0.56 | 71.60 | 2 | 5636096 | 688 | 0.54 | 68.80 | 3 | 5545984 | 677 | 0.53 | 67.70 | 4 | 5308416 | 648 | 0.51 | 64.80 | 5 | 5947392 | 726 | 0.57 | 72.60 | 6 | 5414912 | 661 | 0.52 | 66.10 | 7 | 5431296 | 663 | 0.52 | 66.30 | The total: 44490752 | 5431 | 4.24 | 543.10 | For Read IO thread | bytes | ----------------------------------------------------------------------------------------------------- 0 | 5341184 | 652 | 0.51 | 65.20 | 1 | 5865472 | 716 | 0.56 | 71.60 | 2 | 5636096 | 688 | 0.54 | 68.80 | 3 | 5545984 | 677 | 0.53 | 67.70 | 4 | 5308416 | 648 | 0.51 | 64.80 | 5 | 5947392 | 726 | 0.57 | 72.60 | 6 | 5414912 | 661 | 0.52 | 66.10 | 7 | 5431296 | 663 | 0.52 | 66.30 | ----------------------------------------------------------------------------------------------------- total: 44490752 | 5431 | 4.24 | 543.10 | As Write IO thread | bytes | ----------------------------------------------------------------------------------------------------- 0 | 0 | 0 | 0.00 | 0.00 | 0.000 | N/A | F:\testfile.dat (1000GB) 1 | 0 | 0 | 0.00 | 0.00 | 0.000 | N/A | F:\testfile.dat (1000GB) 2 | 0 | 0 | 0.00 | 0.00 | 0.000 | N/A | F:\testfile.dat (1000GB) 3 | 0 | 0 | 0.00 | 0.00 | 0.000 | N/A | F:\testfile.dat (1000GB) 4 | 0 | 0 | 0.00 | 0.00 | 0.000 | N/A | F:\testfile.dat (1000GB) 5 | 0 | 0 | 0.00 | 0.00 | 0.000 | N/A | F:\testfile.dat (1000GB) 6 | 0 | 0 | 0.00 | 0.00 | 0.000 | N/A | F:\testfile.dat (1000GB) 7 | 0 | 0 | 0.00 | 0.00 | 0.000 | N/A | F:\testfile.dat (1000GB) ----------------------------------------------------------------------------------------------------- total: 0 | 0 | 0.00 | 0.00 | 0.000 | N/A %-ile | ---------------------------------------------- min | 25th | 50th | 75th | 90th | 95th | 99th | 3-nines 4-nines 5-nines 6-nines 7-nines 8-nines max | |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
* DO NOT USE. SQLIO and
SQLIOStress are deprecated and will officially be pulled from download site, if
they are not already. Any personal copies of the tools should be removed.
[转帖]UsingDiskspdforSQLServer --- 改天做下练习 官方文档.的更多相关文章
- 【AutoMapper官方文档】DTO与Domin Model相互转换(下)
写在前面 AutoMapper目录: [AutoMapper官方文档]DTO与Domin Model相互转换(上) [AutoMapper官方文档]DTO与Domin Model相互转换(中) [Au ...
- Spring Cloud官方文档中文版-Spring Cloud Config(下)-客户端等
官方文档地址为:http://cloud.spring.io/spring-cloud-static/Dalston.SR2/#_serving_alternative_formats 文中例子我做了 ...
- ubuntu16.04下snort的安装(官方文档安装)(图文详解)
不多说,直接上干货! 最近为了科研,需要安装和使用Snort. snort的官网 https://www.snort.org/ Snort作为一款优秀的开源主机入侵检测系统,在windows和Linu ...
- ubuntu14.04下snort的安装(官方文档安装)(图文详解)
不多说,直接上干货! 最近为了科研,需要安装和使用Snort. snort的官网 https://www.snort.org/ Snort作为一款优秀的开源主机入侵检测系统,在windows和Linu ...
- vue插件官方文档,做个记录
vue的插件,组件都可以按照这种方式添加 官方文档 https://cn.vuejs.org/v2/guide/plugins.html 做个记录用
- 参考Dubbo3官方文档做的学习笔记
文章目录 概念与架构 2.1 服务发现 Dubbo3官方文档: https://dubbo.apache.org 服务:是 Dubbo 中的核心概念,一个服务代表一组 RPC 方法的集合,服务是面向用 ...
- 【AutoMapper官方文档】DTO与Domin Model相互转换(上)
写在前面 AutoMapper目录: [AutoMapper官方文档]DTO与Domin Model相互转换(上) [AutoMapper官方文档]DTO与Domin Model相互转换(中) [Au ...
- 2DToolkit官方文档中文版打地鼠教程(三):Sprite Collections 精灵集合
这是2DToolkit官方文档中 Whack a Mole 打地鼠教程的译文,为了减少文中过多重复操作的翻译,以及一些无必要的句子,这里我假设你有Unity的基础知识(例如了解如何新建Sprite等) ...
- 【AutoMapper官方文档】DTO与Domin Model相互转换(中)
写在前面 AutoMapper目录: [AutoMapper官方文档]DTO与Domin Model相互转换(上) [AutoMapper官方文档]DTO与Domin Model相互转换(中) [Au ...
随机推荐
- 4-[多进程]-互斥锁、Queue队列、生产者消费者
1.互斥锁 (1)为什么需要互斥锁 进程之间数据不共享,但是共享同一套文件系统,所以访问同一个文件,或同一个打印终端,是没有问题的, 而共享带来的是竞争,竞争带来的结果就是错乱,如下 #并发运行,效率 ...
- 24-[jQuery]-案例
1.仿淘宝导航栏案例 <!DOCTYPE html> <html lang="en"> <head> <meta charset=&quo ...
- idea中xml打开方式变成file,改回来
原文:https://blog.csdn.net/u012903926/article/details/80682885 创建了一个test文件,用的是普通text打开方式,然后你修改文件为test. ...
- 【操作系统】C语言编写的FAT16文件系统
[操作系统]C语言编写的FAT16文件系统 这是操作系统的期末课程设计作业之一,主要功能是在物理内存中虚拟出一个1M大小的FAT16的文件系统,然后把它读入内存中,进行具体的文件操作,具体的实用性不大 ...
- Python 从零搭建 Conf_Web 配置管理平台
作者:Eagle 某船舶行业科技公司,运维工程师,51Reboot学员.通过在51Reboot学习,由运维工程师转至运维开发工程师.完成公司自动化平台的构建,对运维开发有了自己的理解,空闲时间写了这么 ...
- 初识IT行业,人生苦短,我学python
第一次写,我也不知道该怎么写.只有慢慢的去体会大神们的见解与看法. Python是一个较强的脚本语言,而Java是强类型的编程语言.为了更好的入门,我没有去选择强类型语言的Java,而选择使用Pyth ...
- Linux golang使用cgo调用C++标准库问题
我们知道cgo无法直接调用c++方法,但是可以通过c包装c++方法,以达到使用的目的. C++中,我们经常会用到STL.在cgo中,如果要调用STL,需要作如下操作: //cgo LDFLAGS: - ...
- windows上的mysql配置过程
个人电脑的mysql配置,记录下来留作备忘 1. 首先去官网下载最新的mysql安装包,我下的是5.7.25,地址是 https://dev.mysql.com/downloads/windows/ ...
- Dubbo问题处理集合
1 . 启动微服务的时候,报错信息如下: 核心:Can not lock the registry cache file /root/.dubbo/dubbo-registry-127.0.0.1.c ...
- 笔记:《机器学习训练秘籍》——吴恩达deeplearningai微信公众号推送文章
说明 该文为笔者在微信公众号:吴恩达deeplearningai 所推送<机器学习训练秘籍>系列文章的学习笔记,公众号二维码如下,1到15课课程链接点这里 该系列文章主要是吴恩达先生在机器 ...