Need help deciphering top

schwim · 2018-06-22 18:38:32

Hi there everyone!

I've searched and read a bunch but I'm having a problem figuring out what, exactly, top is telling me about the machine I'm running it on.

The machine is an i7 2.66 (listed as 4 core/8 thread) with 16GB of ram running Deb 9.

You can find the top in question here

I've been watching this for the day to see if the fivem process I'm running it taking to much of a toll for the machine.

The process at the top says it's using 32% of CPU. Load at the top says .24/.12 .

Then there's this line:

%Cpu(s): 2.4 us, 0.6 sy, 0.0 ni, 96.8 id, 0.1 wa, 0.0 hi, 0.1 si, 0.0 st

I'm having problems bringing all of this together for an informed picture of what the server load really is.

Could someone help me out by dropping a knowledge bomb on me?

Thanks!

Last edited by schwim (2018-06-22 18:40:18)

iMBeCil · 2018-06-23 06:58:03

^For the top output: see this.

In particular, it seems your cpu is 96.8% not used (at the moment).

As for the question: "what the server load is" is not precise question enough IMHO, which is typical for such questions. For example:
- what should be included in "server load" (i/o load? cpu load? network i/o?
- what time span should we consider? (last second? last minute? last microsecond?)
...

Hope this helps.

schwim · 2018-06-23 12:44:52

HI there imbecil and thanks a bunch for the help!

I do have somewhat of an understanding for the load averages, Reading my topic, I was very unclear in what confused me(not including the line I posted that you explained).

My load averages show .10 to .25, which on a quad core processor, should mean I'm averaging a quarter of that number in capacity (4.0 being 100 computing capacity). The top process, however, shows 30% and that is pretty stable over the 15 minutes. What does that 32% cpu represent? Clearly, not capacity and it doesn't even jive with the capacity of a single core.

I'm confused at the dichotomy between the load and the cpu used on that process.

sleekmason · 2018-06-23 13:25:30

I've been watching this for the day to see if the fivem process I'm running it taking to much of a toll for the machine.

For a better way than top to count performance events and such, you could try perf (performance analysis tools for Linux)

Perf is kernel version dependent, so you want the meta package:

apt-get install linux-perf

For a simple analysis over a period of time, start the process using:

perf stat <program>

in a terminal. When you close the program, the output will show in the terminal window.
Example:

perf stat thunar


 Performance counter stats for 'thunar':

        505.618444      task-clock:u (msec)       #    0.056 CPUs utilized          
                 0      context-switches:u        #    0.000 K/sec                  
                 0      cpu-migrations:u          #    0.000 K/sec                  
             4,098      page-faults:u             #    0.008 M/sec                  
       740,350,986      cycles:u                  #    1.464 GHz                    
       636,657,206      instructions:u            #    0.86  insn per cycle         
       123,804,133      branches:u                #  244.857 M/sec                  
         4,374,851      branch-misses:u           #    3.53% of all branches        

       9.044618080 seconds time elapsed

There are a HUGE number of ways to use perf, including extra unstable programs available in the repos.

I note you said server. You can also start perf on already running instances. See here for details: https://perf.wiki.kernel.org/index.php/Tutorial

PackRat · 2018-06-23 17:11:31

The top process, however, shows 30% and that is pretty stable over the 15 minutes. What does that 32% cpu represent?

That is the tasks share of the amount of CPU use. So its's like 32% of the 2.4% being used (my math may not be precise, but that's how I think of it).

The man page for top has some details about each category that may clear things up.

I like this explanation for load average.

schwim · 2018-06-23 23:57:27

Thank you guys so much for your help! With your help, I've gotten a pretty good idea of what I'm looking at now.

Thank you again!

sleekmason · 2018-06-24 19:45:12

Load averages and CPU utilization (really the % of a time frame the CPU spends on the listed task times) are different. %CPU is likewise a share of time (just want to emphasize this, as it's obviously different than %MEM).

I'm learning all sorts of good stuff here:)

Trying to get a better grip on this. When I use perf to show cpu%, I get different numbers than top.

perf top -nKs pid

   PerfTop:    5880 irqs/sec  kernel:98.9%  exact:  0.0% [3500Hz cycles:pp],  (all, 2 CPUs)
------------------------------------------------------------------------------------------------------

    49.41%         35060        0:swapper        
    25.93%         12967      543:Xorg           
     6.60%          5074     5384:x-www-browser  
     5.60%          3256      771:conky          
     2.53%          1826     7724:x-terminal-emul
     2.11%          1516      725:clipit         
     1.20%           647      741:conky          
     0.57%           513     5398:BgHangManager  
     0.50%           390      424:NetworkManager 
     0.47%           463     8181:perf           
     0.38%           350     8178:perf           
     0.37%           334     2315:irq/17-b43     
     0.37%           275     5400:Timer          
     0.35%           324      818:conky          
     0.34%           224     8186:modprobe       
     0.32%           224     8187:modprobe       
     0.28%           187     8182:sh             
     0.25%           182     7400:x-terminal-emul
     0.24%           202     7694:kworker/1:0    
     0.23%           178      968:ntpd           
     0.22%           165     7448:kworker/0:0    
     0.19%           140     8184:modprobe       
     0.18%           147     7230:kworker/u4:1   
     0.16%           120      825:conky          
     0.15%            77      407:irqbalance     
     0.14%            98      419:dbus-daemon    
     0.11%            92      748:conky          
     0.11%            84     8185:modprobe       
     0.10%            92      824:conky          
     0.10%            55     8183:cat            
     0.08%            49      840:gdbus          
     0.08%            80      820:conky          
     0.07%            49      812:nm-applet      
     0.06%            64      720:tint2          
     0.05%            36      822:conky          
     0.04%            28       14:watchdog/1     
     0.03%            39        7:rcu_sched      
     0.03%            28       15:migration/1    
     0.03%            20     6851:kworker/u4:0   
     0.02%            18       11:watchdog/0

My question is: Why does the cpu% in top show 2.4% for xorg, while the above shows 25.93%?

is top showing the % of the total available, while perf is showing the % within that 2.4%?

If a new thread is necessary, let me know.

Last edited by sleekmason (2018-06-24 19:58:32)

#1 2018-06-22 18:38:32

Need help deciphering top

#2 2018-06-23 06:58:03

Re: Need help deciphering top

#3 2018-06-23 12:44:52

Re: Need help deciphering top

#4 2018-06-23 13:25:30

Re: Need help deciphering top

#5 2018-06-23 17:11:31

Re: Need help deciphering top

#6 2018-06-23 23:57:27

Re: Need help deciphering top

#7 2018-06-24 19:45:12

Re: Need help deciphering top

Board footer