ubuntu下kvm的磁盘io性能优化步骤
1、virsh shutdown wcltest2
2、virsh edit wcltest2
<driver name='qemu' type='qcow2'/>
<source file='/kvm-data/kvm/wcltest2_os_disk.qcow2'/>
<backingStore/>
<target dev='hda' bus='ide'/>
<alias name='ide0-0-0'/>
<address type='drive' controller='0' bus='0' target='0' unit='0'/>
磁盘优化:
<driver name='qemu' type='qcow2' cache='none' io='native'/>
<target dev='vda' bus='virtio'/>
同时删除
<address type='drive' controller='0' bus='0' target='0' unit='0'/>
3、ubuntu下安装sysbench1.0
wget https://github.com/akopytov/sysbench/archive/1.0.zip -O "sysbench-1.0.zip"
unzip sysbench-1.0.zip
apt-get install automake -y
apt-get install libtool -y
./autogen.sh
./configure --without-mysql
备注( --without-mysql 不编译测试mysql的相关环境)
make
make install
4、io测试随机读写
sysbench --test=fileio --num-threads=50 --file-total-size=2G --file-test-mode=rndrw prepare 准备测试
sysbench --test=fileio --num-threads=50 --file-total-size=2G --file-test-mode=rndrw run 开始测试
sysbench --test=fileio --num-threads=50 --file-total-size=2G --file-test-mode=rndrw cleanup 清除测试文件
物理机执行结果:
File operations:
reads/s: 469.90
writes/s: 312.81
fsyncs/s: 1000.35Throughput:
read, MiB/s: 7.34
written, MiB/s: 4.89General statistics:
total time: 10.1041s
total number of events: 18024Latency (ms):
min: 0.00
avg: 27.86
max: 1047.11
95th percentile: 153.02
sum: 502103.21Threads fairness:
events (avg/stddev): 360.4800/49.51
execution time (avg/stddev): 10.0421/0.03
kvm优化前执行结果:File operations:
reads/s: 243.34
writes/s: 162.23
fsyncs/s: 513.40Throughput:
read, MiB/s: 3.80
written, MiB/s: 2.53General statistics:
total time: 10.1048s
total number of events: 9290Latency (ms):
min: 0.00
avg: 54.13
max: 769.09
95th percentile: 297.92
sum: 502877.36Threads fairness:
events (avg/stddev): 185.8000/65.60
execution time (avg/stddev): 10.0575/0.02
kvm优化后结果:
File operations:
reads/s: 438.57
writes/s: 289.51
fsyncs/s: 911.55Throughput:
read, MiB/s: 6.85
written, MiB/s: 4.52General statistics:
total time: 10.1056s
total number of events: 16577Latency (ms):
min: 0.01
avg: 30.30
max: 311.64
95th percentile: 123.28
sum: 502289.90Threads fairness:
events (avg/stddev): 331.5400/28.37
execution time (avg/stddev): 10.0458/0.02
结论:
kvm优化的io随机读写性能达到接近物理机的磁盘io性能