cancel
Showing results for 
Search instead for 
Did you mean: 

Nice value

 
SOLVED
Go to solution
anoop_10
Frequent Advisor

Nice value

Hi ,

We are facing a performance issue with rp3440 server running with HPUX11.11.In top command cpu nice value is showing high and from sar I can see that user % is almost 80.As I have checked with application team they are running some process with lower the priority. I am bit confused with this will making the process priority lower than the normal will affect the system performance.
Please give some inputs Thanks in advance

# sar 5 5

HP-UX scom03 B.11.11 U 9000/800 02/28/09

22:43:14 %usr %sys %wio %idle
22:43:19 71 29 0 0
22:43:24 56 44 0 0
22:43:29 65 35 0 0
22:43:34 77 23 0 0
22:43:39 74 26 0 0

Average 69 31 0 0
# top

System: scom03 Sat Feb 28 22:43:45 2009
Load averages: 24.25, 26.36, 27.59
471 processes: 409 sleeping, 59 running, 3 zombies
Cpu states:
CPU LOAD USER NICE SYS IDLE BLOCK SWAIT INTR SSYS
0 23.89 1.0% 73.3% 25.7% 0.0% 0.0% 0.0% 0.0% 0.0%
1 24.61 0.0% 73.3% 26.7% 0.0% 0.0% 0.0% 0.0% 0.0%
--- ---- ----- ----- ----- ----- ----- ----- ----- -----
avg 24.25 0.0% 74.0% 26.0% 0.0% 0.0% 0.0% 0.0% 0.0%

Memory: 5568404K (5060816K) real, 7483688K (6855696K) virtual, 113220K free Page# 1/14

Thanks
Anoop
7 REPLIES
Solution

Re: Nice value

>will making the process priority lower than the normal will affect the system performance.

I don't see how. Setting nice means what it says, you are being a nice guy and letting other processes get more CPU time because you don't care how long you take.
The overall system performance shouldn't change.
Note: Putting processes in the background typically sets nice. Yours all seem to be niced.

>We are facing a performance issue with rp3440

You need more CPUs or less work.

You are using 25% system, how much swapping are you doing:
swapinfo -tam
anoop_10
Frequent Advisor

Re: Nice value

Thanks for your quick response...

# swapinfo -tam
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 4096 0 4096 0% 0 - 1 /dev/vg00/lvol2
dev 4096 340 3756 8% 0 - 0 /dev/vg00/lvol9
dev 8192 338 7854 4% 0 - 0 /dev/vg00/swap
dev 10224 343 9881 3% 0 - 0 /dev/vgswap/lvol1
reserve - 5534 -5534
memory 6307 1980 4327 31%
total 32915 8535 24380 26% - 0 -
#

Regards
Anoop

Re: Nice value

>memory 6307 1980 4327 31%

This isn't bad. Are you doing this when you have the performance problem?

Also what is the output of "vmstat 1 5"?
Emil Velez
Honored Contributor

Re: Nice value



User mode vs system mode is only a measure of what the programs are doing. A program is in user mode when it is doing math.

A program is in system mode when it does I/O.

unless you rewrite the program you cannot normally change that.

Now you do have a large load average which means that you have 25 processes (or threads) waiting to execute. Imagine a grocery store where the lines are 20 people long to checkout. YOu evidently have a busy system. Now you might be low on memory and vhand has to run to keep freeing up memory or you have too much running on that system.

If response time is ok great otherwise it looks like you need more CPUs.
Anoopkumar
Frequent Advisor

Re: Nice value

Hi Dennis,

Please find the attached file. I do agree with u r statements. I think there is a huge CPU bottleneck and we need to add more CPUs.
In top o/p I can see there is virtual is 100MB free. could you please tell me what is this value and how can I increase it.

Thanks
Anoop

Re: Nice value

>I think there is a huge CPU bottleneck and we need to add more CPUs.

It looks like it. There aren't that many Page Outs.

>In top output I can see there is virtual is 100MB free.

Don't look to top for detailed info. Look at "swapinfo -tam".
anoop_10
Frequent Advisor

Re: Nice value

Thanks for replies...