我们知道评估io性能的三个关键指标为:
1 IOPS
每秒钟处理的IO请求数量。IOPS是随机访问类型业务(OLTP类)很重要的一个参考指标。
2 IO Response Time
IO的响应时间。IO响应时间是从操作系统内核发出一个IO请求到接收到IO响应的时间。因此,IO Response time除了包括磁盘获取数据的时间,还包括了操作系统以及在存储系统内部IO等待的时间。
3 Throughput
吞吐量。这个指标衡量标识了最大的数据传输量。如上说明,这个值在顺序访问或者大数据量访问的情况下会比较重要。尤其在大数据量写的时候。
如何计算IOPS ? 本文通过利用/proc/disksats 的内容来计算磁盘iops,如果是多个磁盘,需要多个变量来标记每个磁盘的累计值和前一秒的值,做减法操作。
首先介绍一下/proc/diskstats 的意义
root@rac1 markbench]# cat /proc/diskstats
1 0 ram0 0 0 0 0 0 0 0 0 0 0 0
1 1 ram1 0 0 0 0 0 0 0 0 0 0 0
3 0 hda 215832 48798 8460369 1779489 10387352 10706454 278983110 129355207 12 30046343 131143587
3 1 hda1 328 1805 2367 1482 24 11 61 1518 0 1499 3000
3 2 hda2 214383 41704 8442154 1766493 10387210 10706256 278981160 129335402 12 30040228 131110596
3 3 hda3 174 1250 1455 920 0 0 0 0 0 110 920
3 4 hda4 4 0 8 9 0 0 0 0 0 9 9
3 5 hda5 925 4006 13977 10447 118 187 1889 18287 0 7409 28734
202 32 xvdc 38 76 912 14 0 0 0 0 0 14 14
202 16 xvdb 40 76 928 17 0 0 0 0 0 17 17
22 0 hdc 18 86 416 39 0 0 0 0 0 35 39
9 0 md0 0 0 0 0 0 0 0 0 0 0 0
这个命令用于显示磁盘、分区和统计信息:hda为整个硬盘的统计信息,hda1为第一个分区的统计信息,hda2为第二个分区的统计信息。
ramdisk设备为通过软件将RAM当做硬盘来使用的一项技术。
[root@rac1 markbench]# cat /sys/block/hda/hda2/stat
214428 41704 8443522 1767344 10431899 10785067 282040104 130435808 9 30120018 132201965
/proc/diskstats文件比/sys/block/hda/hda2/stat文件多3列,从左至右分别对应主设备号,次设备号和设备名称。后续的11列的意义在这两个文件里是相同的,除了第9列,所有的域都是从启动时的累积值。
第1 列 读完成次数,成功完成读的总次数。
(number of issued reads. This is the total number of reads completed successfully.)
第2 列 合并读完成次数, 第6 列合并写完成次数。为了效率可能会合并相邻的读和写。从而两次4K的读在它最终被处理到磁盘上之前可能会变成一次8K的读,才被计数(和排队),因此只有一次I/O操作。该值使你知道这样的操作有多频繁。
(number of reads merged)
第3 列 读扇区的次数,成功读过的扇区总次数。
(number of sectors read. This is the total number of sectors read successfully.)
第4 列 读花费的毫秒数,这是所有读操作所花费的毫秒数(用__make_request()到end_that_request_last()测量)。
(number of milliseconds spent reading. This is the total number of milliseconds spent by all reads (as measured from __make_request() to end_that_request_last()).)
第5 列 写完成次数,成功写完成的总次数。
(number of writes completed. This is the total number of writes completed successfully.)
第6 列 合并写完成次数
(number of writes merged Reads and writes which are adjacent to each other may be merged for efficiency. Thus two 4K reads may become one 8K read before it is ultimately handed to the disk, and so it will be counted (and queued) as only one I/O. This field lets you know how often this was done.)
第7 列 写扇区次数,成功写扇区总次数。
(number of sectors written. This is the total number of sectors written successfully.)
第8 列 写操作花费的毫秒数,这是所有写操作所花费的毫秒数(用__make_request()到end_that_request_last()测量)。
(number of milliseconds spent writing This is the total number of milliseconds spent by all writes (as measured from __make_request() to end_that_request_last()).)
第9 列 正在处理的输入/输出请求数-I/O的当前进度,只有这个域应该是0。当请求被交给适当的request_queue_t时增加和请求完成时减小。
(number of I/Os currently in progress. The only field that should go to zero. Incremented as requests are given to appropriate request_queue_t and decremented as they finish.)
第10 列 输入/输出操作花费的毫秒数,花在I/O操作上的毫秒数,这个域会增长只要field 9不为0。
(number of milliseconds spent doing I/Os. This field is increased so long as field 9 is nonzero.)
第11 列 输入/输出操作花费的加权毫秒数 花在I/O操作上的毫秒数,在每次I/O开始,I/O结束,I/O合并时这个域都会增加。这可以给I/O完成时间和存储那些可以累积的提供一个便利的测量标准。
(number of milliseconds spent doing I/Os. This field is incremented at each I/O start, I/O completion, I/O merge, or read of these stats by the number of I/Os in progress (field 9) times the number of milliseconds spent doing I/O since the last update of this field. This can provide an easy measure of both I/O completion time and the backlog that may be accumulating.)
统计的iops 的脚本如下:
#!/bin/bash
uplrio=0
uplwio=0
updrio=0
updwio=0
while true ; do
lrio=$(grep hda2 /proc/diskstats | awk '{print $4}')
lwio=$(grep hda2 /proc/diskstats | awk '{print $8}')
llrio=$(echo $lrio - $uplrio | bc)
llwio=$(echo $lwio - $uplwio | bc)
iops=$(echo "$llrio + $llwio " | bc)
echo "iops:$iops Data_Read $llrio Data_Write $llwio "
uplrio=$lrio
uplwio=$lwio
sleep 1
done
效果展示:
[root@rac1 markbench]# sh iops.sh
iops:10348521 Data_Read 214045 Data_Write 10134476 --第一个因为是累计值-0的结果，所以比较大,可以忽略。
iops:322 Data_Read 0 Data_Write 322
iops:596 Data_Read 0 Data_Write 596
iops:589 Data_Read 0 Data_Write 589
iops:615 Data_Read 0 Data_Write 615
iops:599 Data_Read 0 Data_Write 599
iops:455 Data_Read 0 Data_Write 455
iops:533 Data_Read 0 Data_Write 533
iops:516 Data_Read 0 Data_Write 516
iops:214 Data_Read 0 Data_Write 214
iops:544 Data_Read 0 Data_Write 544