P550小机IO很慢，dd裸设备IO却是正常的
一台P550小机的IO很慢，系统是AIX6.1，新机器，两块盘做了raid1，没跑业务，下面是IO的情况
在/下：
# time dd if=/dev/zero of=./test.dmp bs=1M count=1000
1000+0 records in.
1000+0 records out.
real 1m30.42s
user 0m0.02s
sys 0m3.33s
topas结果：
Disk Busy% KBPS TPS KB-R ART MRT KB-W AWT MWT AQW AQD
hdisk1 99.0 10.1K 201.0 0.0 0.0 7.1 10.1K 7.7 117.2 0.0 0.0
hdisk0 79.0 10.1K 201.0 0.0 0.0 9.5 10.1K 7.8 63.2 0.0 0.0
# lsvg -l rootvg
rootvg:
LV NAME TYPE LPs PPs PVs LV STATE MOUNT POINT
hd5 boot 1 2 2 closed/syncd N/A
hd6 paging 128 256 2 open/syncd N/A
hd8 jfs2log 1 2 2 open/syncd N/A
hd4 jfs2 20 40 2 open/syncd /
hd2 jfs2 32 64 2 open/syncd /usr
hd9var jfs2 32 64 2 open/syncd /var
hd3 jfs2 40 80 2 open/syncd /tmp
hd1 jfs2 16 32 2 open/syncd /home
hd10opt jfs2 20 40 2 open/syncd /opt
hd11admin jfs2 1 2 2 open/syncd /admin
lg_dumplv sysdump 16 16 1 open/syncd N/A
livedump jfs2 1 2 2 open/syncd /var/adm/ras/livedump
lv_test jfs 10 10 2 closed/syncd N/A
dd裸设备：
# time dd if=/dev/zero of=/dev/rlv_test bs=1M count=1000
1000+0 records in.
1000+0 records out.
real 0m8.70s
user 0m0.01s
sys 0m0.60s
topas结果：
Disk Busy% KBPS TPS KB-R ART MRT KB-W AWT MWT AQW AQD
hdisk1 94.0 57.8K 462.0 0.0 0.0 13.2 57.8K 6.3 101.2 0.0 0.0
hdisk0 89.0 57.7K 461.0 0.0 0.0 17.8 57.7K 5.7 117.6 0.0 0.0
昨天在这个操作系统上去dd一块新盘，io也是上面的结果。后面又在新盘上重新装个操作系统，IO又一切正常。然后上面dd文件系统的话，io很慢，dd 裸设备又是正常的，那这是不是证明了硬件没问题，是操作系统的问题吧，请问这是什么原因造成的？操作系统有哪个参数限制了吗?
下面是我另一台同样配置的机器的测试结果：
在/下：
# time dd if=/dev/zero of=./test.dmp bs=1M count=1000
1000+0 records in.
1000+0 records out.
real 0m9.64s
user 0m0.01s
sys 0m3.11s
topas结果：
Disk Busy% KBPS TPS KB-R ART MRT KB-W AWT MWT AQW AQD
hdisk0 99.0 94.5K 140.0 0.0 0.0 31.4 94.5K 59.2 336.4 9.8 0.0
hdisk1 87.0 95.5K 133.0 0.0 0.0 99.6 95.5K 51.3 2.0K 8.0 0.0
dd裸设备：
# time dd if=/dev/zero of=/dev/rlv_test bs=1M count=1000
1000+0 records in.
1000+0 records out.
real 0m9.78s
user 0m0.01s
sys 0m0.62s
问题解决
smitty chgsys
两个参数
HIGH water mark for pending write I/Os per file
LOW water mark for pending write I/Os per file
这两个值改成默认就好了，因为做了双机，按照以前的经验和aix5L上一样都是改成33和24，在aix6.1上保持默认就行
下面是IBM对这两个参数的解释
vaule:
aix6:
The default value for maxpout is 8193, and minpout is 4096
http://publib.boulder.ibm.com/in ... /disk_io_pacing.htm
aix5:
The default value for both the maxpout and minpout parameters is 0, which means that the I/O pacing feature is disabled.
http://publib.boulder.ibm.com/in ... /disk_io_pacing.htm
the best performance:
The best range for the maxpout and minpout parameters depends on the CPU speed and the I/O system. I/O pacing works well if the value of the maxpout parameter is equal to or greater than the value of the j2_nPagesPerWriteBehindCluster parameter. For example, if the value of the maxpout parameter is equal to 64 and the minpout parameter is equal to 32, there are at most 64 pages in I/O state and 2 I/Os before blocking on the next write.
The default tuning parameters are as follows:
Parameter Default Value
j2_nPagesPerWriteBehindCluster 32
j2_nBufferPerPagerDevice 512
在/下：
# time dd if=/dev/zero of=./test.dmp bs=1M count=1000
1000+0 records in.
1000+0 records out.
real 1m30.42s
user 0m0.02s
sys 0m3.33s
topas结果：
Disk Busy% KBPS TPS KB-R ART MRT KB-W AWT MWT AQW AQD
hdisk1 99.0 10.1K 201.0 0.0 0.0 7.1 10.1K 7.7 117.2 0.0 0.0
hdisk0 79.0 10.1K 201.0 0.0 0.0 9.5 10.1K 7.8 63.2 0.0 0.0
# lsvg -l rootvg
rootvg:
LV NAME TYPE LPs PPs PVs LV STATE MOUNT POINT
hd5 boot 1 2 2 closed/syncd N/A
hd6 paging 128 256 2 open/syncd N/A
hd8 jfs2log 1 2 2 open/syncd N/A
hd4 jfs2 20 40 2 open/syncd /
hd2 jfs2 32 64 2 open/syncd /usr
hd9var jfs2 32 64 2 open/syncd /var
hd3 jfs2 40 80 2 open/syncd /tmp
hd1 jfs2 16 32 2 open/syncd /home
hd10opt jfs2 20 40 2 open/syncd /opt
hd11admin jfs2 1 2 2 open/syncd /admin
lg_dumplv sysdump 16 16 1 open/syncd N/A
livedump jfs2 1 2 2 open/syncd /var/adm/ras/livedump
lv_test jfs 10 10 2 closed/syncd N/A
dd裸设备：
# time dd if=/dev/zero of=/dev/rlv_test bs=1M count=1000
1000+0 records in.
1000+0 records out.
real 0m8.70s
user 0m0.01s
sys 0m0.60s
topas结果：
Disk Busy% KBPS TPS KB-R ART MRT KB-W AWT MWT AQW AQD
hdisk1 94.0 57.8K 462.0 0.0 0.0 13.2 57.8K 6.3 101.2 0.0 0.0
hdisk0 89.0 57.7K 461.0 0.0 0.0 17.8 57.7K 5.7 117.6 0.0 0.0
昨天在这个操作系统上去dd一块新盘，io也是上面的结果。后面又在新盘上重新装个操作系统，IO又一切正常。然后上面dd文件系统的话，io很慢，dd 裸设备又是正常的，那这是不是证明了硬件没问题，是操作系统的问题吧，请问这是什么原因造成的？操作系统有哪个参数限制了吗?
下面是我另一台同样配置的机器的测试结果：
在/下：
# time dd if=/dev/zero of=./test.dmp bs=1M count=1000
1000+0 records in.
1000+0 records out.
real 0m9.64s
user 0m0.01s
sys 0m3.11s
topas结果：
Disk Busy% KBPS TPS KB-R ART MRT KB-W AWT MWT AQW AQD
hdisk0 99.0 94.5K 140.0 0.0 0.0 31.4 94.5K 59.2 336.4 9.8 0.0
hdisk1 87.0 95.5K 133.0 0.0 0.0 99.6 95.5K 51.3 2.0K 8.0 0.0
dd裸设备：
# time dd if=/dev/zero of=/dev/rlv_test bs=1M count=1000
1000+0 records in.
1000+0 records out.
real 0m9.78s
user 0m0.01s
sys 0m0.62s
问题解决
smitty chgsys
两个参数
HIGH water mark for pending write I/Os per file
LOW water mark for pending write I/Os per file
这两个值改成默认就好了，因为做了双机，按照以前的经验和aix5L上一样都是改成33和24，在aix6.1上保持默认就行
下面是IBM对这两个参数的解释
vaule:
aix6:
The default value for maxpout is 8193, and minpout is 4096
http://publib.boulder.ibm.com/in ... /disk_io_pacing.htm
aix5:
The default value for both the maxpout and minpout parameters is 0, which means that the I/O pacing feature is disabled.
http://publib.boulder.ibm.com/in ... /disk_io_pacing.htm
the best performance:
The best range for the maxpout and minpout parameters depends on the CPU speed and the I/O system. I/O pacing works well if the value of the maxpout parameter is equal to or greater than the value of the j2_nPagesPerWriteBehindCluster parameter. For example, if the value of the maxpout parameter is equal to 64 and the minpout parameter is equal to 32, there are at most 64 pages in I/O state and 2 I/Os before blocking on the next write.
The default tuning parameters are as follows:
Parameter Default Value
j2_nPagesPerWriteBehindCluster 32
j2_nBufferPerPagerDevice 512