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.

iostat 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 345 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-2shows 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 requests than others. A consequence of this could be to move certain parts of a database from the busiest disks to less used ones.

Example: A-2 SUN Solaris iostat output

#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 iostat command is different for every UNIX variant. It also offers many different options and parameters. Refer to your system documentation and the iostat manpage for more information.

System Activity Report (SAR)

The System Activity Report (SAR) provides a quick way to tell if a system is I/O bound. 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 347. 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 all disks, both internal and external. Here, too, the first line represents the average since boot time and should be discarded.

346DS8000 Series: Concepts and Architecture

Page 368
Image 368
IBM DS8000 manual System Activity Report SAR, Example A-2 SUN Solaris iostat output