如何进行磁盘的I/O测试

如何进行磁盘的I/O测试

https://wiki.mikejung.biz/Benchmarking

译文有增删

1 Linux 常见基准测试工具

FIO

Sysbench

Phoronix Test Suite (moved to it’s own
page)

IOzone

Ioping

UnixBench

Google’s Perfkit Benchmarker

2 FIO

fio算是比较老的io测试工具了，作者是Jens Axboe。

主页https://www.thomas-krenn.com/en/wiki/Fio

fio的安装

我的测试环境有centos7 redhat5 因此要下载两个版本的fio包

redhat5 版本的需要去归档站点去下载

http://archives.fedoraproject.org/pub/archive/epel/

下载后的安装 rpm -ivh
fio-1.57-1.el5.x86_64.rpm

centos7 则可以去很多镜像站点下载，站点列表可以查询

https://admin.fedoraproject.org/mirrormanager/mirrors/EPEL

如果配置了epel 的yum 源，直接yum install -y fio

fio测试选项和例子

示例 使用fio测试磁盘的随机写

fio –name=randwrite –ioengine=libaio –iodepth=1
–rw=randwrite –bs=4k

 
–direct=0 –size=512M –numjobs=8 –runtime=240 –group_reporting

–rw=randwrite 指定随机写

–direct=0 指定使用buffered IO 还是direct IO。如果使用buffer，注意总大小不要超过物理内存

–numjobs=8 启动8个进程

–size=512M 每个进程写512M

–group_reporting  将多个进程的统计结果进行聚合，更易阅读

–runtime=240 持续运行时间4分钟

–bs=4k 
–blocksize=4k (default)

–ioengine=libaio 可用的io引擎可以使用 fio –enghelp查看(fio-2.2.8)

–iodepth=1 缺省 测试ssd磁盘时，可以扩大到32，通常4就够了 

The iodepth
option defines the amount of IO units that will continue to hammer a file with
requests during the test.

避免使用buffer的方法是指定–direct=1 或者使用比物理内存大一倍的写文件

fio结果的重点关注项目：

iops=1416        iops = 1416

95.00th=[    2]  95%的IO在2毫秒完成

util=99.62%      设备繁忙程度

示例 使用fio测试磁盘的随机读

fio –name=randread –ioengine=libaio
–iodepth=16 –rw=randread –bs=4k

 
–direct=0 –size=512M –numjobs=8 –runtime=240 –group_reporting

fio 帮助

man fio

fio -h

fio -v

fio 命令行中写上选项，也可以将命令行选项写入一个参数文件job file中

fio 安装后，自带了一些样例

# rpm -ql -p fio-1.57-1.el5.x86_64.rpm

/usr/bin/fio

/usr/bin/fio_generate_plots

/usr/share/doc/fio-1.57

/usr/share/doc/fio-1.57/COPYING

/usr/share/doc/fio-1.57/HOWTO

/usr/share/doc/fio-1.57/README

/usr/share/doc/fio-1.57/REPORTING-BUGS

/usr/share/doc/fio-1.57/examples

/usr/share/doc/fio-1.57/examples/1mbs_clients

/usr/share/doc/fio-1.57/examples/aio-read

……

aio-read 参数文件，修改一下可以马上使用

# fio aio-read.bak

file1: (g=0): rw=randread,
bs=128K-128K/128K-128K, ioengine=libaio, iodepth=4

file2: (g=0): rw=randread,
bs=128K-128K/128K-128K, ioengine=libaio, iodepth=32

file3: (g=0): rw=randread,
bs=128K-128K/128K-128K, ioengine=libaio, iodepth=8

file4: (g=0): rw=randread,
bs=128K-128K/128K-128K, ioengine=libaio, iodepth=16

fio 1.57

……

fio 输出结果的一些注释

  
io     Number of megabytes of I/O
performed.

  
bw     Average data rate
(bandwidth).

  
runt   Threads run time.

  
slat   Submission  latency 
minimum, maximum, average and standard deviation. This

         
is the time it took to submit the I/O.

  
clat   Completion latency minimum,
maximum, average and standard deviation.  
This

         
is the time between submission and completion.

  
bw     Bandwidth  minimum, 
maximum,  percentage  of aggregate bandwidth received, average and
standard deviation.

  
The group statistics show:

         
io     Number of megabytes I/O
performed.

         
aggrb  Aggregate bandwidth of threads
in the group.

         
minb   Minimum average bandwidth a
thread saw.

         
maxb   Maximum average bandwidth a
thread saw.

         
mint   Shortest runtime of threads
in the group.

         
maxt   Longest runtime of threads
in the group.

   Finally,
disk statistics are printed with reads first:

         
ios    Number of I/Os performed by
all groups.

         
merge  Number of merges in the I/O
scheduler.

         
ticks  Number of ticks we kept the
disk busy.

         
io_queue

                 Total time spent in the disk queue.

         
util   Disk utilization.

示例 随机读直接IO

fio –name=randread –ioengine=libaio
–iodepth=16 –rw=randread –bs=4k –direct=1 –size=1G –numjobs=8
–runtime=240 –group_reporting

csdn上一个翻译成中文的帖子 http://blog.csdn.net/yuesichiu/article/details/8722417 没耐心看英文的估计中文的也不会看完。

3 sysbench

sysbench可以测试cpu oltp mysql fileio 网上用来测试mysql的文章比较多。

sysbench的安装

在centos 中，可以从epel直接安装。安装时需要mysql / postgresql

源文中的安装方法

实际安装方法

rpm -ivh
http://archives.fedoraproject.org/pub/archive/epel/epel-release-latest-5.noarch.rpm

or

rpm -ivh
http://dl.fedoraproject.org/pub/epel/epel-release-latest-7.noarch.rpm

yum install sysbench

sysbench的I/O测试

先掌握sysbench的语法

# sysbench

Missing required command argument.

Usage:

  sysbench [general-options]…
–test= [test-options]… command

General options:

 
–num-threads=N            number
of threads to use [1]

 
–max-requests=N           limit
for total number of requests [10000]

 
–max-time=N               limit
for total execution time in seconds [0]

 
–forced-shutdown=STRING   amount
of time to wait after –max-time before forcing shutdown [off]

 
–thread-stack-size=SIZE   size of
stack per thread [32K]

 
–init-rng=[on|off]        initialize random number generator [off]

 
–test=STRING              test to
run

 
–debug=[on|off]           print
more debugging info [off]

 
–validate=[on|off]        perform
validation checks where possible [off]

 
–help=[on|off]            print help
and exit

 
–version=[on|off]         print
version and exit

Compiled-in tests:

 
fileio – File I/O test

  cpu
– CPU performance test

 
memory – Memory functions speed test

 
threads – Threads subsystem performance test

 
mutex – Mutex performance test

 
oltp – OLTP test

Commands: prepare run cleanup help version

来一些命令

# run a random write test with
”–file-extra-flags=direct” 随机写

sysbench
–test=fileio –file-total-size=4G –file-num=2 –file-test-mode=rndwr
–max-time=240 –max-requests=0 –file-block-size=8K –num-threads=4
–file-extra-flags=direct

  prepare

  run

  cleanup

sysbench
–test=fileio –file-total-size=1G –file-num=10 –file-test-mode=rndwr
–max-time=240 –max-requests=0 –file-block-size=8K –file-extra-flags=direct

或者

随机写

sysbench –test=fileio –file-total-size=2G
–file-num=64 prepare

sysbench –test=fileio –file-total-size=2G
–file-test-mode=rndwr –max-time=240 –max-requests=0 –file-block-size=4K
–file-num=64 –num-threads=1 run

for each in 1 4 8 16 32 64; do

sysbench –test=fileio –file-total-size=2G
–file-test-mode=rndwr –max-time=240 –max-requests=0 –file-block-size=4K
–file-num=64 –num-threads=$each run;

sleep 10; done

随机读

sysbench –test=fileio –file-total-size=2G
–file-test-mode=rndrd –max-time=240 –max-requests=0 –file-block-size=4K
–file-num=64 –num-threads=1 run

for each in 1 4 8 16 32 64; do sysbench
–test=fileio –file-total-size=2G –file-test-mode=rndrd –max-time=240
–max-requests=0 –file-block-size=4K –file-num=64 –num-threads=$each run;
sleep 10; done

测试结果中的几个重点：

一个虚拟机(openstack)的测试结果

[root@v-yfb-cos72-sql05 test]# sysbench
–test=fileio –file-total-size=4G –file-num=2 –file-test-mode=rndwr –max-time=240
–max-requests=0 –file-block-size=8K –num-threads=4 run

sysbench 0.4.12:  multi-threaded system evaluation benchmark

Running the test with following options:

Number of threads: 4

Extra file open flags: 0

2 files, 2Gb each

4Gb total file size

Block size 8Kb

Number of random requests for random IO: 0

Read/Write ratio for combined random IO
test: 1.50

Periodic FSYNC enabled, calling fsync()
each 100 requests.

Calling fsync() at the end of test,
Enabled.

Using synchronous I/O mode

Doing random write test

Threads started!

Time limit exceeded, exiting…

(last message repeated 3 times)

Done.

Operations performed:  0 Read, 14433721 Write, 288270 Other =
14721991 Total

Read 0b 
Written 110.12Gb  Total
transferred 110.12Gb  (469.84Mb/sec)
<—- 60140.13 * 8k = 469.84

60140.13 Requests/sec executed <—- 重点1 IOPS

Test execution summary:

   
total time:                         
240.0015s

   
total number of events:             
14433721

   
total time taken by event execution: 66.3879

   
per-request statistics:

        
min:                                  0.00ms

        
avg:                                 
0.00ms

        
max:                                 
2.05ms

        
approx.  95 percentile:               0.01ms <—- 重点2 每个请求的响应时间

Threads fairness:

   
events (avg/stddev):          
3608430.2500/38671.29

   
execution time (avg/stddev):  
16.5970/0.16

另外一个虚拟机(VMware esx)的测试结果

[root@dbsrvb test]# sysbench –test=fileio
–file-total-size=4G –file-num=2 –file-test-mode=rndwr –max-time=240
–max-requests=0 –file-block-size=8K –num-threads=4 run

sysbench 0.4.12:  multi-threaded system evaluation benchmark

Running the test with following options:

Number of threads: 4

Extra file open flags: 0

2 files, 2Gb each

4Gb total file size

Block size 8Kb

Number of random requests for random IO: 0

Read/Write ratio for combined random IO
test: 1.50

Periodic FSYNC enabled, calling fsync()
each 100 requests.

Calling fsync() at the end of test,
Enabled.

Using synchronous I/O mode

Doing random write test

Threads started!

FATAL: Failed to fsync file! file:
-1667944680 errno = 9 ()

FATAL: Incorrect file discovered in request

(last message repeated 2 times)

Done.

Operations performed:  0 Read, 10503 Write, 208 Other = 10711 Total

Read 0b 
Written 82.055Mb  Total
transferred 82.055Mb  (21.993Mb/sec)

 2815.05 Requests/sec executed <—- 重点1 IOPS

Test execution summary:

   
total time:                         
3.7310s

   
total number of events:             
10503

   
total time taken by event execution: 0.9682

   
per-request statistics:

        
min:                              
   0.01ms

        
avg:                                 
0.09ms

        
max:                                
17.76ms

        
approx.  95 percentile:               0.22ms <—- 重点2 比上面的服务器要差一些

Threads fairness:

   
events (avg/stddev):          
2625.7500/377.73

   
execution time (avg/stddev):  
0.2421/0.02

另外一个虚拟机(笔记本上的VirtualBox)的测试结果

[root@node1 test]# sysbench –test=fileio
–file-total-size=4G –file-num=2 –file-test-mode=rndwr –max-time=240
–max-requests=0 –file-block-size=8K –num-threads=4 run

sysbench 0.4.12:  multi-threaded system evaluation benchmark

Running the test with following options:

Number of threads: 4

Extra file open flags: 0

2 files, 2Gb each

4Gb total file size

Block size 8Kb

Number of random requests for random IO: 0

Read/Write ratio for combined random IO
test: 1.50

Periodic FSYNC enabled, calling fsync()
each 100 requests.

Calling fsync() at the end of test,
Enabled.

Using synchronous I/O mode

Doing random write test

Threads started!

Time limit exceeded, exiting…

(last message repeated 3 times)

Done.

Operations performed:  0 Read, 635100 Write, 12701 Other = 647801
Total

Read 0b 
Written 4.8454Gb  Total
transferred 4.8454Gb  (20.673Mb/sec)

 2646.16 Requests/sec executed <— Virtual Box 的 IOPS 比VMware稍微低一点

Test execution summary:

   
total time:                         
240.0086s

   
total number of events:             
635100

   
total time taken by event execution: 2.7852

   
per-request statistics:

        
min:                                 
0.00ms

        
avg:                                  0.00ms

        
max:                                 
3.75ms

        
approx.  95 percentile:               0.01ms <— 这个就比VMware快

Threads fairness:

   
events (avg/stddev):          
158775.0000/3440.64

   
execution time (avg/stddev):  
0.6963/0.02

如果测试硬件性能，要在物理服务器上进行对比。

为了避免从内存读取数据，测试文件要大于内存的两倍。

sysbench  的mysql测试

##Create Database

mysql ‐u root ‐ppassword ‐e ”CREATE DATABASE sysbench;”

##Create User

mysql ‐u root ‐ppassword ‐e ”CREATE USER ’sysbench’@’localhost’ IDENTIFIED BY
’password’;”

##Grant Access

mysql ‐u root ‐ppassword ‐e ”GRANT ALL PRIVILEGES ON *.* TO ’sysbench’@’localhost’
IDENTIFIED BY ’password’;”

或者

mysql> create database sysbench;

mysql> create user ’sysbench’@’%’
identified by ’sysbench’ ;

mysql> select host,user from mysql.user;

+———–+———–+

| host     
| user      |

+———–+———–+

| %        
| sysbench  |

| localhost | mysql.sys |

| localhost | root      |

+———–+———–+

3 rows in set (0.00 sec)

mysql> grant all privileges on *.* to
’sysbench’@’%’ ;

mysql> flush privileges;

数据库环境准备完毕后，准备sysbench命令

sysbench –test=oltp –db-driver=mysql
–oltp-table-size=1000000

–mysql-db=sysbench –mysql-user=sysbench
–mysql-password=sysbench

prepare

自动产生的测试表的样子

mysql> desc sbtest

   
-> ;

+——-+——————+——+—–+———+—————-+

| Field | Type             | Null | Key | Default |
Extra          |

+——-+——————+——+—–+———+—————-+

| id   
| int(10) unsigned | NO   | PRI |
NULL    | auto_increment |

| k    
| int(10) unsigned | NO   | MUL |
0       |                |

| c    
| char(120)        | NO   |    
|         |                |

| pad  
| char(60)         | NO   |    
|         |                |

+——-+——————+——+—–+———+—————-+

4 rows in set (0.00 sec)

mysql>

mysql> select * from sbtest limit 5;

+—-+—+—+—————————————————-+

| id | k | c | pad                                               
|

+—-+—+—+—————————————————-+

|  1
| 0 |   |
qqqqqqqqqqwwwwwwwwwweeeeeeeeeerrrrrrrrrrtttttttttt |

|  2
| 0 |   |
qqqqqqqqqqwwwwwwwwwweeeeeeeeeerrrrrrrrrrtttttttttt |

|  3
| 0 |   |
qqqqqqqqqqwwwwwwwwwweeeeeeeeeerrrrrrrrrrtttttttttt |

|  4
| 0 |   |
qqqqqqqqqqwwwwwwwwwweeeeeeeeeerrrrrrrrrrtttttttttt |

|  5
| 0 |   |
qqqqqqqqqqwwwwwwwwwweeeeeeeeeerrrrrrrrrrtttttttttt |

+—-+—+—+—————————————————-+

5 rows in set (0.00 sec)

for each in 1 4 8 16 32 64; do

sysbench –test=oltp –db-driver=mysql
–oltp-table-size=10000000

–mysql-db=sysbench –mysql-user=sysbench
–mysql-password=sysbench

–max-time=240 –max-requests=0
–oltp-read-only –oltp-skip-trx

–oltp-nontrxmode=select
–num-threads=$each run;

sleep 10; done

sysbench –test=oltp –db-driver=mysql
–oltp-table-size=10000000

–mysql-db=sysbench –mysql-user=sysbench
–mysql-password=sysbench

–max-time=240 –max-requests=0
–oltp-read-only –oltp-skip-trx

–oltp-nontrx-mode=select –num-threads=2
run;

4 Phoronix 测试套件

Phoronix 测试套件简称pts, 从官网http://phoronix-test-suite.com/?k=downloads下载安装包 phoronix-test-suite-7.0.1.tar.gz

pts的安装依赖php-cli 等多个包

tar xvf phoronix-test-suite_6.2.2.tar.gz

cd phoronix-test-suite/

./install-sh

/usr/bin/phoronix-test-suite
list-available-suites

/usr/bin/phoronix-test-suite
list-available-tests

phoronix-test-suite info pts/disk

phoronix-test-suite install pts/iozon

phoronix-test-suite install pts/disk

pts/disk包括了fio iozone aio-stress等等多个工具，pts太过庞大，install时因为网速原因，耗时较长。这里只使用一个iozone做个测试。

使用pts套件运行iozone

[root@node1 ~]# phoronix-test-suite run
pts/iozone-1.8.0

IOzone 3.405:

   
pts/iozone-1.8.0

   
Disk Test Configuration

       
1: 4Kb

       
2: 64Kb

       
3: 1MB

       
4: Test All Options

       
Record Size: 1

       
1: 512MB

       
2: 2GB

       
3: 4GB

        4: 8GB

       
5: Test All Options

       
File Size: 4

       
1: Write Performance

       
2: Read Performance

       
3: Test All Options

       
Disk Test: 1

Phoronix Test Suite v6.6.0

System Information

Hardware:

Processor: Intel Core i7-3537U @ 2.49GHz (1
Core), Motherboard: Oracle VirtualBox v1.2, Chipset: Intel 440FX- 82441FX PMC,
Memory: 4096MB, Disk: 72GB VBOX HDD, Graphics: InnoTek VirtualBox, Audio: Intel
82801AA AC 97 Audio, Network: Intel 82540EM Gigabit

Software:

OS: Oracle Linux Server 7.3, Kernel:
4.1.12-61.1.18.el7uek.x86_64 (x86_64), Compiler: GCC 4.8.5 20150623,
File-System: xfs, System Layer: KVM VirtualBox

   
Would you like to save these test results (Y/n): y

   
Enter a name to save these results under: tmp

   
Enter a unique name to describe this test run / configuration: tmp

If desired, enter a new description below
to better describe this result set / system configuration under test.

Press ENTER to proceed without changes.

Current Description: KVM VirtualBox testing
on Oracle Linux Server 7.3 via the Phoronix Test Suite.

New Description:

IOzone 3.405:

   
pts/iozone-1.8.0 [Record Size: 4Kb – File Size: 8GB – Disk Test: Write
Performance]

   
Test 1 of 1

   
Estimated Trial Run Count:    3

   
Estimated Time To Completion: 22 Minutes (04:14 CDT)

       
Started Run 1 @ 03:53:01

       
Started Run 2 @ 04:11:11

       
Started Run 3 @ 04:22:17  [Std.
Dev: 108.27%]

       
Started Run 4 @ 04:26:15  [Std.
Dev: 83.68%]

       
Started Run 5 @ 04:29:54  [Std.
Dev: 69.61%]

       
Started Run 6 @ 04:33:32  [Std.
Dev: 60.77%]

   
Test Results:

       
8.509765625

       
14.8359375

       
69.3681640625

       
74.802734375

       
74.82421875

       
76.93359375

   
Average: 53.21 MB/s

   
Do you want to view the results in your web browser (Y/n): n 

   
Would you like to upload the results to OpenBenchmarking.org (Y/n): n

单独运行iozone

# find / -name iozone

/var/lib/phoronix-test-suite/installed-tests/pts/iozone-1.8.0/iozone3_405/src/current/iozone

/var/lib/phoronix-test-suite/installed-tests/pts/iozone-1.8.0/iozone

# cd
/var/lib/phoronix-test-suite/installed-tests/pts/iozone-1.8.0/

# export LOG_FILE=1.log

# ./iozone -Ra -g 8G -i 0 -b iozone.xls

参考

http://phoronix-test-suite.com/documentation/phoronix-test-suite.html#GettingStarted

https://wiki.mikejung.biz/Phoronix_Test_Suite

5 IOzone 的安装和使用

官网 http://iozone.org/

官方帮助文档 IOzone_msword_98.doc

下载安装包  [Stable
tarball]  iozone3_465.tar

安装过程

 
150  mkdir test_iozone

 
151  mv iozone3_465.tar
test_iozone/

 
152  cd test_iozone/

 
153  ls

 
154  tar -xf iozone3_465.tar

 
155  cd iozone3_465/src/current/

 
156  ls

 
157  make

 
158  uname -a

 
159  make linux-AMD64

 
160  yum provides cc

 
161  yum install -y gcc

 
162  make linux-AMD64

 
163  ll

 
164  ls -ort

 
165  date

 
166  iozone

 
167  ./iozone

 
168  ./iozone -h

 
169  ./iozone -Ra -g 8G -i 0 -b
iozone.xls 这个输出到xls文件

运行命令

 ./iozone -Ra -g 8G -i 0  这个输出到屏幕

iozone产生的结果比较直观地看到不同块大小和不同文件大小对IO的影响。

7 UnixBench 的安装和使用

apt‐get install make gcc

wget
https://byte‐unixbench.googlecode.com/files/UnixBench6.1.3.tgz

tar zxvf UnixBench6.1.3.tgz

cd UnixBench

make

./Run ‐i 1

  download  UnixBench6.1.3.tgz from

   
https://code.google.com/archive/p/byte-unixbench/downloads

 
255  tar -xzf UnixBench6.1.3.tgz

 
256  ls

 
257  cd UnixBench

 
258  ls

 
259  make

 
260  ls -ort

 
261  ./Run -i 1

 
262  which perl

 
263  yum install -y perl

[root@v-yfb-cos72-sql05 UnixBench]# ./Run -i 1

make all

make[1]: Entering directory
`/root/UnixBench’

Checking distribution of files

./pgms 
exists

./src 
exists

./testdir 
exists

./tmp 
exists

./results 
exists

make[1]: Leaving directory
`/root/UnixBench’

sh: 3dinfo: command not found

  
#    #  #   
#  #  #   
#          #####   ###### 
#    #   ####  
#    #

  
#    #  ##  
#  #   # 
#           #    # 
#       ##   # 
#    #  #    #

  
#    #  # # 
#  #    ##            #####   #####  
# #  #  #      
######

  
#    #  #  #
#  #   
##            #    # 
#       #  # # 
#       #    #

  
#    #  #  
##  #   # 
#           #    # 
#       #   ## 
#    #  #    #

   
####   #    # 
#  #    #         
#####   ######  #   
#   ####   #    #

  
Version 5.1.3                     
Based on the Byte Magazine Unix Benchmark

  
Multi-CPU version                 
Version 5 revisions by Ian Smith,

                                     
Sunnyvale, CA, USA

  
January 13, 2011                  
johantheghost at yahoo period com

……

========================================================================

  
BYTE UNIX Benchmarks (Version 5.1.3)

  
System: v-yfb-cos72-sql05.novalocal: GNU/Linux

  
OS: GNU/Linux — 3.10.0-327.el7.x86_64 — #1 SMP Thu Nov 19 22:10:57 UTC
2015

  
Machine: x86_64 (x86_64)

  
Language: en_US.utf8 (charmap=”UTF-8”, collate=”UTF-8”)

  
CPU 0: Intel Core Processor (Haswell, no TSX) (4600.0 bogomips)

         
x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET

  
CPU 1: Intel Core Processor (Haswell, no TSX) (4600.0 bogomips)

         
x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET

  
06:01:21 up 78 days,  3:34,  1 user, 
load average: 0.11, 0.04, 0.05; runlevel 3

————————————————————————

Benchmark Run: Thu May 04 2017 06:01:21 –
06:08:05

2 CPUs in system; running 1 parallel copy
of tests

Dhrystone 2 using register variables       29114714.8 lps   (10.0 s, 1 samples)

Double-Precision Whetstone                     3660.1 MWIPS (9.8 s, 1
samples)

Execl Throughput                               3227.4 lps   (29.1 s, 1 samples)

File Copy 1024 bufsize 2000 maxblocks        962272.0 KBps  (30.0 s, 1 samples)

File Copy 256 bufsize 500 maxblocks          256736.0 KBps  (30.0 s, 1 samples)

File Copy 4096 bufsize 8000 maxblocks       2537028.0 KBps  (30.0 s, 1 samples)

Pipe Throughput                             1282585.0 lps   (10.0 s, 1 samples)

Pipe-based Context Switching                 284899.6 lps   (10.0 s, 1 samples)

Process Creation                               9563.0 lps   (30.0 s, 1 samples)

Shell Scripts (1 concurrent)                   6315.2 lpm   (60.0 s, 1 samples)

Shell Scripts (8 concurrent)                   1360.1 lpm   (60.0 s, 1 samples)

System Call Overhead                        2215239.6 lps   (10.0 s, 1 samples)

System Benchmarks Index Values               BASELINE       RESULT   
INDEX

Dhrystone 2 using register variables         116700.0   29114714.8  
2494.8

Double-Precision Whetstone                       55.0       3660.1   
665.5

Execl Throughput                                 43.0       3227.4   
750.6

File Copy 1024 bufsize 2000 maxblocks          3960.0     962272.0  
2430.0

File Copy 256 bufsize 500 maxblocks            1655.0     256736.0  
1551.3

File Copy 4096 bufsize 8000 maxblocks          5800.0    2537028.0  
4374.2

Pipe Throughput                               12440.0    1282585.0  
1031.0

Pipe-based Context Switching                   4000.0     284899.6   
712.2

Process Creation                                126.0       9563.0   
759.0

Shell Scripts (1 concurrent)                     42.4       6315.2  
1489.4

Shell Scripts (8 concurrent)                      6.0       1360.1  
2266.8

System Call Overhead                          15000.0    2215239.6  
1476.8

                                                 
                 ========

System Benchmarks Index Score                                        1400.8

————————————————————————

Benchmark Run: Thu May 04 2017 06:08:05 –
06:14:49

2 CPUs in system; running 2 parallel copies
of tests

Dhrystone 2 using register variables       58393583.7 lps   (10.0 s, 1 samples)

Double-Precision Whetstone                     7356.6 MWIPS (9.8 s, 1
samples)

Execl Throughput                               7037.2 lps   (29.9 s, 1 samples)

File Copy 1024 bufsize 2000 maxblocks       1368114.0 KBps  (30.0 s, 1 samples)

File Copy 256 bufsize 500 maxblocks          390873.0 KBps  (30.0 s, 1 samples)

File Copy 4096 bufsize 8000 maxblocks       3596554.0 KBps  (30.0 s, 1 samples)

Pipe Throughput                             2688283.4 lps   (10.0 s, 1 samples)

Pipe-based Context Switching                 572723.9 lps   (10.0 s, 1 samples)

Process Creation                              22877.0 lps   (30.0 s, 1 samples)

Shell Scripts (1 concurrent)                   9545.3 lpm   (60.0 s, 1 samples)

Shell Scripts (8 concurrent)                   1391.3 lpm   (60.0 s, 1 samples)

System Call Overhead                        3284659.7 lps   (10.0 s, 1 samples)

System Benchmarks Index Values               BASELINE       RESULT   
INDEX

Dhrystone 2 using register variables         116700.0   58393583.7  
5003.7

Double-Precision Whetstone                       55.0       7356.6  
1337.6

Execl Throughput                                 43.0       7037.2   1636.6

File Copy 1024
bufsize 2000 maxblocks          3960.0    1368114.0  
3454.8

File Copy 256 bufsize 500 maxblocks            1655.0     390873.0  
2361.8

File Copy 4096 bufsize 8000 maxblocks          5800.0    3596554.0  
6201.0

Pipe Throughput                               12440.0    2688283.4  
2161.0

Pipe-based Context Switching                   4000.0     572723.9  
1431.8

Process Creation                                126.0      22877.0  
1815.6

Shell Scripts (1 concurrent)                     42.4      
9545.3   2251.2

Shell Scripts (8 concurrent)                      6.0       1391.3  
2318.8

System Call Overhead                          15000.0    3284659.7  
2189.8

                                                                  
========

System Benchmarks Index Score                                        2394.8

[root@v-yfb-cos72-sql05 UnixBench]#

[root@node1 UnixBench]# ./Run -i 1

make all

make[1]: Entering directory
`/root/UnixBench’

Checking distribution of files

./pgms 
exists

./src 
exists

./testdir 
exists

./tmp 
exists

./results 
exists

make[1]: Leaving directory
`/root/UnixBench’

sh: 3dinfo: command not found

  
#    #  #   
#  #  #   
#          #####   ###### 
#    #   ####  
#    #

  
#    #  ##  
#  #   # 
#           #    # 
#       ##   # 
#    #  #    #

  
#    #  # # 
#  #    ##            #####   #####  
# #  #  #      
######

  
#    #  #  #
#  #   
##            #    # 
#       #  # # 
#       #    #

  
#    #  #  
##  #   # 
#           #    # 
#       #   ## 
#    # 
#    #

   
####   #    # 
#  #    #         
#####   ######  #   
#   ####   #    #

  
Version 5.1.3                     
Based on the Byte Magazine Unix Benchmark

  
Multi-CPU version                 
Version 5 revisions by Ian Smith,

            
                         Sunnyvale, CA, USA

  
January 13, 2011                  
johantheghost at yahoo period com

…… 这个工具不单单是测试I/O。

8 ioping 的安装和使用

下载 ioping-0.8.tar.gz

wget
https://ioping.googlecode.com/files/ioping‐0.8.tar.gz

tar zxvf ioping‐0.8.tar.gz

cd ioping‐0.8/

make

make install

./ioping ‐c 1000 /home/

##To run ioping in current directory

./ioping ‐c 1000 .

## run ioping with direct I/O

./ioping ‐D ‐c 1000 .

# ./ioping -D -c 10 .

4.0 KiB from . (xfs /dev/mapper/centos-root): request=1
time=2.7 ms

4.0 KiB from . (xfs /dev/mapper/centos-root): request=2
time=330 us

4.0 KiB from . (xfs /dev/mapper/centos-root): request=3
time=382 us

4.0 KiB from . (xfs /dev/mapper/centos-root): request=4
time=383 us

4.0 KiB from . (xfs /dev/mapper/centos-root): request=5
time=451 us

4.0 KiB from . (xfs /dev/mapper/centos-root): request=6
time=280 us

4.0 KiB from . (xfs /dev/mapper/centos-root): request=7
time=435 us

4.0 KiB from . (xfs /dev/mapper/centos-root): request=8
time=292 us

4.0 KiB from . (xfs /dev/mapper/centos-root): request=9
time=241 us

4.0 KiB from . (xfs /dev/mapper/centos-root): request=10 time=309 us

— . (xfs /dev/mapper/centos-root) ioping statistics —

10 requests completed in 9.0 s, 1.7
k iops, 6.8 MiB/s

min/avg/max/mdev = 241 us / 576 us / 2.7 ms / 697 us

注意上面的时间单位 microsecond(us)=1/1，000，000秒  ms (millisecond)=1/1000秒

下面引用褚霸总结的性能必知数字 《利用新硬件提升数据库性能.pptx》

 nanoseconds
(ns) = 1/1，000，000，000 秒

9 PerfKitBenchmarker

参考 https://wiki.mikejung.biz/Google_PerfKitBenchmarker

这个没有做测试。

apt-get install git python-pip
python-dev unzip openjdk-7-jre

git clone
https://github.com/GoogleCloudPlatform/PerfKitBenchmarker.git

cd PerfKitBenchmarker/

pip install
-r requirements.txt

10 AIX 下磁盘 I/O 性能分析

https://www.ibm.com/developerworks/cn/aix/library/1203_weixy_aixio/index.html

表 1. 常见磁盘类型及其 IOPS

 http://en.wikipedia.org/wiki/IOPS