Total CPU usage during peak periods has been above 80% .

[redhat7012]

Hi,

How to check the CPU Utilization in HP-UX ?

How to decrease the CPU utilization in HP-UX Box ?

Regards,

Aparna.

[redhat7012]

Load averages: 0.16, 0.16, 0.18
647 processes: 598 sleeping, 48 running, 1 zombie
Cpu states:
CPU   LOAD   USER   NICE    SYS   IDLE  BLOCK  SWAIT   INTR   SSYS
 0    0.19   5.8%   1.2%  14.1%  78.9%   0.0%   0.0%   0.0%   0.0%
 1    0.23  11.7%   0.4%   9.3%  78.5%   0.0%   0.0%   0.0%   0.0%
 2    0.22  11.3%   2.8%   6.6%  79.3%   0.0%   0.0%   0.0%   0.0%
 3    0.21  11.7%   0.2%   9.5%  78.5%   0.0%   0.0%   0.0%   0.0%
 4    0.19   4.2%   0.4%  13.5%  81.9%   0.0%   0.0%   0.0%   0.0%
 5    0.15   7.4%   0.8%   7.4%  84.5%   0.0%   0.0%   0.0%   0.0%
 6    0.18   2.8%   1.4%   8.7%  87.1%   0.0%   0.0%   0.0%   0.0%
 7    0.18   6.2%   0.8%  11.7%  81.3%   0.0%   0.0%   0.0%   0.0%
 8    0.16   4.4%   0.4%  11.7%  83.5%   0.0%   0.0%   0.0%   0.0%
 9    0.15   9.5%   0.2%   7.6%  82.7%   0.0%   0.0%   0.0%   0.0%
10    0.14   7.2%   1.4%   7.6%  83.9%   0.0%   0.0%   0.0%   0.0%
11    0.14   2.4%   1.0%   8.0%  88.7%   0.0%   0.0%   0.0%   0.0%
12    0.14   6.0%   0.0%   9.1%  84.9%   0.0%   0.0%   0.0%   0.0%
13    0.14   3.2%   0.8%   5.2%  90.9%   0.0%   0.0%   0.0%   0.0%
14    0.12   2.8%   0.0%  10.1%  87.1%   0.0%   0.0%   0.0%   0.0%
15    0.14   3.2%   0.4%   8.0%  88.5%   0.0%   0.0%   0.0%   0.0%
16    0.16   2.4%   0.6%   5.8%  91.3%   0.0%   0.0%   0.0%   0.0%
17    0.18  17.9%   0.6%   9.9%  71.6%   0.0%   0.0%   0.0%   0.0%
18    0.12   8.7%   0.2%   9.5%  81.5%   0.0%   0.0%   0.0%   0.0%
19    0.12  10.3%   0.4%   6.8%  82.5%   0.0%   0.0%   0.0%   0.0%
---   ----  -----  -----  -----  -----  -----  -----  -----  -----
avg   0.16   7.0%   0.8%   8.9%  83.3%   0.0%   0.0%   0.0%   0.0%

Memory: 8000564K (3682588K) real, 11763240K (5233084K) virtual, 32495796K free  Page# 1/50

CPU TTY    PID USERNAME PRI NI   SIZE    RES STATE    TIME %WCPU  %CPU COMMAND
 4   ?     121 oracle   238 20  7724M  4892K run      1:21 71.88 71.66 oraclePRDUSG01
14   ?   28094 oracle   218 20   151M   132M run    624:21 24.58 24.54 tnslsnr
 0   ?    8937 oracle   148 20  7722M  3868K sleep  108:33 23.21 23.17 oraclePRDUSG01
 5   ?    3798 oracle   148 20  7723M  5468K sleep    0:02  9.80  8.77 ora_j000_PRDUSG01
15   ?   26151 oracle   154 20  7722M  3868K sleep    1:35  5.63  5.62 oraclePRDUSG01
12   ?   18253 oper     148 20  7722M  4112K sleep    1:13  4.02  4.01 oraclePRDUSG01
 9   ?   29937 oracle   152 20   186M   139M run      1:47  2.78  2.77 StealthCollect@tvolga-PRDUSG01@quest_sc_mw
 5   ?    5289 oracle   152 20  7681M 25844K run      0:00  4.17  2.48 StealthCollect@tvolga-PRDUSG01@quest_sc_collector
 3   ?   24032 root     -16 20 91044K 65924K run   1313:15  1.55  1.55 midaemon
10 pts/7 23025 oper     154 24  8224K  4484K sleep    0:01  1.10  1.10 itkwish
17   ?    3524 root     152 20 52760K 16008K run   6503:39  0.96  0.96 nco_m_ssmagent.bin
 9   ?   18945 oracle   154 20  7722M  3868K sleep    0:30  0.95  0.94 oraclePRDUSG01

Is the C.P.U Utilization O.k here ?

Any Suggestions.

We have got the alert when it reached 80% .

[Ken Grabowski]

Hi redhat,

Utilization by itself is not an issue! 80% used still means 20% is not being used. Since these seems to be a pattern, and 80% is your peak utilization, then you are probably OK. Has there been any complaints about system performance?

According to your print out, your 20 processors are averaging 83% idle.  Of course that's a moment in time print out and you need to be looking at your usage over a period of time.  Do you have Glance or Glance Plus Pak with measureware?

With the X-windows gpm version of glance you can open up the Global report and watch the "Pri Queue" and "CPU % **bleep**" fields. If CPU% is near 100% and "Pri Queue" is greater than three, then you are probably encountering a CPU bottleneck. If it is not, then you are simply using the system resources.

If you have Measurware or Glance Plus Pak which includes measureware, you can use the extract command to pull out historical performance data. Usually it is configured to save between 3 and 6 months of data.  Using Measurware you would get the GBL_CPU_TOTAL_UTIL for the CPU_%_Cum and GBL_PRI_QUEUE for the Pri Queue, and look for the same information.

You have a 20 CPU server and there is a good chance you have Glance Plus Pak. But if you don't you can use tools like sar and vmstat to collect historical data and analyze it.  Most of the time a 30 day period will give you a good idea of system load. But depending on what those Oracle databases are used for, and what kind of industry you are in, you many want to know quarter end and year end loads too.  You normally configure a system to handle your worst case load not your daily average. So knowing where those load peaks are, is very important.

Don't forget to watch disk IO too!

[Dennis Handly]

>How to check the CPU Utilization in HP-UX?

top and ps are one way but you need to do this when you get the alert, during the peak.

>How to decrease the CPU utilization in HP-UX Box?

What  HP-UX version do you have?  Also what does model(1) return?

Well, you can always add more memory and CPUs.  Or upgrade your machine to a faster one.

Otherwise you would have to tune your application.  Or run less processes.