测试方法:
两台dell机器
操作系统: Red Hat Enterprise Linux Server release 5.3 (Tikanga)
#uname -a
Linux oratest2 2.6.18-128.el5 #1 SMP Wed Dec 17 11:41:38 EST 2008 x86_64 x86_64 x86_64 GNU/Linux
主要测试AOE设备的响应时间,通过一个内存盘进行测试,由于内存盘本身是没有延迟的,故产生的延迟基本上都是网络产生的,看这种情况下,AOE的延迟是多少。
在/dev/ram0上建一个AOE设备:
./vblade 9 0 eth2 /dev/ram0
然后在另一个机器上运行:
#iopress 0 1 60 8192 e9.0 O_DIRECT
iopress v1.0
Author: TangCheng
Copyright (C) 2008-2009 Alibaba.com Limited
Start to read data test...
Large file is supported,file e9.0 size is 16 Mbytes,blocksize is 8192.
IO count is 391887,Times is 60.00seconds,IOPS is 6531,mean response time is 0.15ms,Cpu is 0.86%
可以看到响应时间为0.15ms,这时的网络流量基本上为50Mbytes/s。
#sar -n DEV 1 2000
Linux 2.6.18-128.el5 (oratest2) 09/24/2010
09:34:12 AM IFACE rxpck/s txpck/s rxbyt/s txbyt/s rxcmp/s txcmp/s rxmcst/s
09:34:13 AM lo 1.00 1.00 52.00 52.00 0.00 0.00 0.00
09:34:13 AM eth2 53090.00 53090.00 3185404.00 56273404.00 0.00 0.00 0.00
09:34:13 AM eth1 0.00 0.00 0.00 0.00 0.00 0.00 0.00
09:34:13 AM eth0 6.00 4.00 512.00 452.00 0.00 0.00 0.00
09:34:13 AM eth3 0.00 0.00 0.00 0.00 0.00 0.00 0.00
当开两个并发,把网络流量压到90Mbytes时,延迟上升到0.17ms,基本上还是不错的。
#iopress 0 2 60 8192 e9.0 O_DIRECT
iopress v1.0
Author: TangCheng
Copyright (C) 2008-2009 Alibaba.com Limited
Start to read data test...
Large file is supported,file e9.0 size is 16 Mbytes,blocksize is 8192.
IO count is 717736,Times is 60.00seconds,IOPS is 11962,mean response time is 0.17ms,Cpu is 2.03%
09:41:50 AM IFACE rxpck/s txpck/s rxbyt/s txbyt/s rxcmp/s txcmp/s rxmcst/s
09:41:51 AM lo 0.00 0.00 0.00 0.00 0.00 0.00 0.00
09:41:51 AM eth2 92306.00 92314.00 5538364.00 97850844.00 0.00 0.00 0.00
09:41:51 AM eth1 0.00 0.00 0.00 0.00 0.00 0.00 0.00
09:41:51 AM eth0 5.00 5.00 346.00 1066.00 0.00 0.00 0.00
09:41:51 AM eth3 0.00 0.00 0.00 0.00 0.00 0.00 0.00
09:41:51 AM IFACE rxpck/s txpck/s rxbyt/s txbyt/s rxcmp/s txcmp/s rxmcst/s
09:41:52 AM lo 3.00 3.00 158.00 158.00 0.00 0.00 0.00
09:41:52 AM eth2 92482.00 92474.00 5548924.00 98020444.00 0.00 0.00 0.00
09:41:52 AM eth1 0.00 0.00 0.00 0.00 0.00 0.00 0.00
09:41:52 AM eth0 5.00 5.00 346.00 1066.00 0.00 0.00 0.00
09:41:52 AM eth3 0.00 0.00 0.00 0.00 0.00 0.00 0.00
09:41:52 AM IFACE rxpck/s txpck/s rxbyt/s txbyt/s rxcmp/s txcmp/s rxmcst/s
09:41:53 AM lo 0.00 0.00 0.00 0.00 0.00 0.00 0.00
09:41:53 AM eth2 76484.00 76492.00 4589044.00 81077984.00 0.00 0.00 0.00
09:41:53 AM eth1 0.00 0.00 0.00 0.00 0.00 0.00 0.00
09:41:53 AM eth0 4.00 4.00 276.00 964.00 0.00 0.00 0.00
09:41:53 AM eth3 0.00 0.00 0.00 0.00 0.00 0.00 0.00
由此可见AOE的延迟要大大低于iscsi。
阅读(1303) | 评论(0) | 转发(0) |