Measure CPU usage for a process on Linux?

On modern multi-core CPUs, monitoring CPU usage of individual cores for specific processes is essential for performance analysis and identifying system bottlenecks. This article demonstrates how to measure CPU core usage for processes on Linux using various command-line tools.

Monitoring CPU Usage with top Command

The top command is the most common tool for monitoring system resources. By default, it shows aggregate CPU usage, but you can view per-core statistics.

To display individual CPU core usage, run top and press 1 to toggle per-core view:

$ top

Then press 1 to see individual core statistics:

top - 10:38:00 up 45 min, 3 users, load average: 0.26, 0.14, 0.15
Tasks: 23 total, 1 running, 22 sleeping, 0 stopped, 0 zombie
%Cpu0 : 5.0 us, 4.0 sy, 0.0 ni, 91.0 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st
%Cpu1 : 4.7 us, 4.0 sy, 0.0 ni, 89.3 id, 2.0 wa, 0.0 hi, 0.0 si, 0.0 st
MiB Mem: 7957.6 total, 6407.9 free, 541.9 used, 1007.8 buff/cache
MiB Swap: 0.0 total, 0.0 free, 0.0 used, 7185.1 avail Mem

 PID USER  PR NI  VIRT   RES   SHR S %CPU %MEM    TIME+ COMMAND
   1 root  20  0  3900  3016  2728 S  0.0  0.0   0:00.02 bash

The %CPU column shows per-process CPU usage, while the top section displays individual core utilization. The values represent: us (user), sy (system), ni (nice), id (idle), and wa (I/O wait).

Using mpstat for Detailed Core Statistics

The mpstat command provides comprehensive per-core CPU statistics. First, install the sysstat package:

$ sudo apt-get install sysstat

Use -P ALL to display statistics for all CPU cores:

$ mpstat -P ALL

Linux 5.10.147+ (cs) 12/19/2022 _x86_64_ (2 CPU)

10:43:33 AM  CPU   %usr  %nice  %sys  %iowait %irq  %soft  %steal  %guest %gnice %idle
10:43:33 AM  all   4.67   0.00  3.53   0.64   0.00   0.06    0.04    0.00   0.00  91.06
10:43:33 AM    0   4.63   0.00  3.36   0.05   0.00   0.04    0.04    0.00   0.00  91.87
10:43:33 AM    1   4.70   0.00  3.70   1.24   0.00   0.07    0.04    0.00   0.00  90.24

For continuous monitoring, use interval and count parameters:

$ mpstat -P ALL 2 5

This displays statistics every 2 seconds for 5 iterations.

Process-Specific Monitoring with perf

The perf command provides detailed performance profiling capabilities. Install it using:

$ sudo apt-get install linux-tools-common

To monitor a specific process by PID, use perf stat:

$ perf stat -p 6565

Performance counter stats for process id '6565':

     1.07 msec task-clock        # 0.003 CPUs utilized
      569      context-switches  # 0.529 K/sec
        0      CPU-migrations    # 0.000 K/sec
       26      page-faults       # 0.024 K/sec
22,547,697    cycles            # 2.108 GHz
 8,262,105    instructions      # 0.37  insn per cycle
 1,981,482    branches          # 1846.695 M/sec
    16,532    branch-misses     # 0.83% of all branches

1.477977927 seconds time elapsed

The CPUs utilized field shows the fraction of available CPU cores used by the process. For per-core analysis, combine perf with taskset to bind processes to specific cores.

Comparison of Tools

Conclusion

Measuring CPU core usage helps identify performance bottlenecks and optimize workload distribution across multi-core systems. The top command provides real-time interactive monitoring, mpstat offers comprehensive system-wide statistics, and perf delivers detailed process-specific profiling capabilities.