IBM DS6000 Series ユーザーズマニュアル

DS6000 Series: Concepts and Architecture

The output reports the following:

򐂰

The %tm_act column indicates the percentage of the measured interval time that the 
device was busy.

򐂰

The Kbps column shows the average data rate, read and write data combined, of this 
device.

򐂰

The tps column shows the transactions per second. Note that an I/O transaction can have 
a variable transfer size. This field may also appear higher than would normally be 
expected for a single physical disk device.

򐂰

The Kb_read and Kb_wrtn columns show the total amount of data read and written.

 can also be issued for continuous monitoring with a given number of iterations and a 

monitoring period. It will then print a report like that in Example A-1 on page 301 for every 
period, with the values calculated for exactly this period. In most cases this mode is more 
useful, because bottlenecks mostly appear only during peak times and are not reflected in an 
overall average. Be aware that the first in the series of reports represents the average since 
boot and should be discarded. 

Example A-2 shows an iostat report from SUN Solaris. You see an example of a device that 
appears to be very busy (sd1). The r/s column shows 124.3 reads per second; the %b column 
shows 90 percent busy. The svc_t column, however, shows a service time of 15.7 ms, still 
quite reasonable for 124 I/Os per second. Depending on the application layout, this report 
could lead to the conclusion that the I/O load of this system is unbalanced. Some disks get a 
lot more I/O request than others. A consequence of this could be to move certain parts of a 
database from the busiest disks to less used ones.

#iostat -x
extended disk statistics
disk  r/s   w/s   Kr/s   Kw/s wait actv svc_t %w %b
fd0

 0.0   0.0

  0.0    0.0  0.0  0.0   0.0  0  0

sd1 124.3  14.5 3390.9  399.7  0.0  2.0  15.7  0 90
sd2   0.7   0.4   13.9

  4.0  0.0  0.0   7.8  0  1

sd3   0.4   0.5

  2.5

  3.8  0.0  0.1   8.1  0  1

sd6   0.0   0.0

  0.0

  0.0  0.0  0.0   5.8  0  0

sd8   0.3   0.2

  9.4

  9.6  0.0  0.0   8.6  0  1

sd9   0.7   1.3   12.4   21.3  0.0  0.0   5.2  0  3

The implementation of the 

 command is different for every UNIX variant. It also offers 

many different options and parameters. Refer to your system documentation and the 

 

man page for more information.

The 

 (SAR) provides a quick way to tell if a system is 

. SAR 

has numerous options, providing paging, TTY, CPU busy, and many other statistics.

One way you can run sar is by specifying a sampling interval and the number of times you 
want it to run. 

This is shown in Example A-3 on page 303. It displays CPU usage information, sampled five 
times with a two second interval. To check whether a system is I/O bound, the important 
column to look at is %wio. The %wio indicates the time spent waiting on I/O from 

both internal and external. Here, too, the first line represents the average since boot time and 
should be discarded.