HP-UX sar -d 分析
#sar -d 1 10
HP-UX sapprd01 B.11.31 U ia64 11/09/11
10:01:03 device %busy avque r+w/s blks/s avwait avserv
disk10 100.00 0.50 2244 36190 0.00 1.58
disk10 100.00 0.50 2438 39046 0.00 1.40
disk10 100.00 0.50 2288 36468 0.00 1.55
disk10 100.00 0.50 2326 38061 0.00 1.44
disk10 100.00 0.50 2818 45188 0.00 1.23
disk10 100.00 0.50 2489 39857 0.00 1.36
disk10 100.00 0.50 2290 36945 0.00 1.50
disk10 100.00 0.50 2612 41842 0.00 1.28
disk10 100.00 0.50 3043 48642 0.00 1.09
disk10 100.00 0.50 2414 38626 0.00 1.43
Average disk10 100.00 0.50 2496 40089 0.00 1.37
disk10 %busy指标持续100%,平均2496次rw/s,40089 blks/s 平均每个I/O大小是8k(40089/2496*512/1024=8k)
其中关键I/O瓶颈性能指标的平均值avque=0.5和avserv=1.37,两个指标的值都很低,完全在正常范围内.
以下是相关性能指标的说明:
device device file name of the device
%busy avg of amount of time during sample period that device was busy
avque avg of how many I/Os in HP-UX are waiting to go to the device
r+w/s avg number of read and write blocks transferred per second
blks/s avg size of tranfers (in 512 byte blocks) per second
avwait avg of how long I/O's are queued in HP-UX (waiting to go - in mSec)
avserv avg of "wire time" (specifically: HBA to device and back - in mSec)
As a general recommedation, good ranges are:
avque < 1.0, but greater values are not a problem
avserv < 20, higher values indicate device / SAN bottleneck
%busy就是在采样周期内disk一直在忙于I/O,并不能说明I/O性能有问题,只要指标avserv <20就说明存储没有瓶颈,在EVA上的分析也可以证明这一点.
这时候我们继续给这个disk增加I/O压力直到avserv >20才说明磁盘出现瓶颈.
#sar -d 1 10
HP-UX sapprd01 B.11.31 U ia64 11/09/11
10:01:03 device %busy avque r+w/s blks/s avwait avserv
disk10 100.00 0.50 2244 36190 0.00 1.58
disk10 100.00 0.50 2438 39046 0.00 1.40
disk10 100.00 0.50 2288 36468 0.00 1.55
disk10 100.00 0.50 2326 38061 0.00 1.44
disk10 100.00 0.50 2818 45188 0.00 1.23
disk10 100.00 0.50 2489 39857 0.00 1.36
disk10 100.00 0.50 2290 36945 0.00 1.50
disk10 100.00 0.50 2612 41842 0.00 1.28
disk10 100.00 0.50 3043 48642 0.00 1.09
disk10 100.00 0.50 2414 38626 0.00 1.43
Average disk10 100.00 0.50 2496 40089 0.00 1.37
disk10 %busy指标持续100%,平均2496次rw/s,40089 blks/s 平均每个I/O大小是8k(40089/2496*512/1024=8k)
其中关键I/O瓶颈性能指标的平均值avque=0.5和avserv=1.37,两个指标的值都很低,完全在正常范围内.
以下是相关性能指标的说明:
device device file name of the device
%busy avg of amount of time during sample period that device was busy
avque avg of how many I/Os in HP-UX are waiting to go to the device
r+w/s avg number of read and write blocks transferred per second
blks/s avg size of tranfers (in 512 byte blocks) per second
avwait avg of how long I/O's are queued in HP-UX (waiting to go - in mSec)
avserv avg of "wire time" (specifically: HBA to device and back - in mSec)
As a general recommedation, good ranges are:
avque < 1.0, but greater values are not a problem
avserv < 20, higher values indicate device / SAN bottleneck
%busy就是在采样周期内disk一直在忙于I/O,并不能说明I/O性能有问题,只要指标avserv <20就说明存储没有瓶颈,在EVA上的分析也可以证明这一点.
这时候我们继续给这个disk增加I/O压力直到avserv >20才说明磁盘出现瓶颈.
编辑回复