Linux中如何计算特定CPU使用率

/proc/stat

kernel/system statistics.  Varies with architecture. 

Common entries include:

     user nice system idle iowait  irq  softirq steal guest guest_nice

cpu  4705 356  584    3699   23    23     0       0     0        0

cpu0 1393280 32966 572056 13343292 6130 0 17875 0 23933 0

   The amount of time, measured in units of USER_HZ

   (1/100ths of a second on most architectures, use

   sysconf(_SC_CLK_TCK) to obtain the right value), that

   the system ("cpu" line) or the specific CPU ("cpuN"

   line) spent in various states:

   user   (1) Time spent in user mode.

   nice   (2) Time spent in user mode with low priority

          (nice).

   system (3) Time spent in system mode.

   idle   (4) Time spent in the idle task.  This value

          should be USER_HZ times the second entry in the

          /proc/uptime pseudo-file.

   iowait (since Linux 2.5.41)

          (5) Time waiting for I/O to complete.  This

          value is not reliable, for the following rea?\

          sons:

          1. The CPU will not wait for I/O to complete;

             iowait is the time that a task is waiting for

             I/O to complete.  When a CPU goes into idle

             state for outstanding task I/O, another task

             will be scheduled on this CPU.

          2. On a multi-core CPU, the task waiting for I/O

             to complete is not running on any CPU, so the

             iowait of each CPU is difficult to calculate.

          3. The value in this field may decrease in cer?\

             tain conditions.

   irq (since Linux 2.6.0-test4)

          (6) Time servicing interrupts.

   softirq (since Linux 2.6.0-test4)

          (7) Time servicing softirqs.

   steal (since Linux 2.6.11)

          (8) Stolen time, which is the time spent in

          other operating systems when running in a virtu?\

          alized environment

   guest (since Linux 2.6.24)

          (9) Time spent running a virtual CPU for guest

          operating systems under the control of the Linux

          kernel.

   guest_nice (since Linux 2.6.33)

          (10) Time spent running a niced guest (virtual

          CPU for guest operating systems under the con?\

          trol of the Linux kernel).

Total CPU time since boot = user+nice+system+idle+iowait+irq+softirq+steal

Total CPU Idle time since boot = idle + iowait

Total CPU usage time since boot = Total CPU time since boot - Total CPU Idle time since boot

Total CPU percentage = Total CPU usage time since boot/Total CPU time since boot * 100%

CPU percentage between t1 and t2 = (（total2-total1）-（idle2-idle1））/（total2-total1）* 100%

#!/bin/bash

# by Paul Colby (http://colby.id.au), no rights reserved ;)

PREV_TOTAL=0

PREV_IDLE=0

while true; do

  # Get the total CPU statistics, discarding the 'cpu ' prefix.

  CPU=(`sed -n 's/^cpu\s//p' /proc/stat`)

  IDLE=${CPU[3]} # Just the idle CPU time.

  # Calculate the total CPU time.

  TOTAL=0

  for VALUE in "${CPU[@]}"; do

    let "TOTAL=$TOTAL+$VALUE"

  done

  # Calculate the CPU usage since we last checked.

  let "DIFF_IDLE=$IDLE-$PREV_IDLE"

  let "DIFF_TOTAL=$TOTAL-$PREV_TOTAL"

  let "DIFF_USAGE=(1000*($DIFF_TOTAL-$DIFF_IDLE)/$DIFF_TOTAL+5)/10"

  echo -en "\rCPU: $DIFF_USAGE%  \b\b"

  # Remember the total and idle CPU times for the next check.

  PREV_TOTAL="$TOTAL"

  PREV_IDLE="$IDLE"

  # Wait before checking again.

  sleep 1

done