Configure kernel tick time

By default, RHEL uses a tickless kernel that does not interrupt idle CPUs, reducing power usage and enabling deep sleep states on newer processors. RHEL also provides a dynamic tickless option for latency-sensitive workloads, such as high-performance and real-time computing.

About this task

By default, the dynamic tickless option is disabled. You can use the cpu-partitioning TuneD profile to enable the dynamic tickless option for cores specified as isolated_cores.

Procedure

  1. To enable dynamic tickless behavior in certain cores, specify those cores on the kernel command line with the nohz_full parameter. For example, on a 16 core system, enable the nohz_full=1-15 kernel option: 
    # grubby --update-kernel=ALL --args="nohz_full=1-15"

    This enables dynamic tickless behavior on cores 1 through 15, moving all timekeeping to the only unspecified core (core 0).

  2. When the system boots, manually move the rcu threads to the non-latency-sensitive core, in this case core 0: 
    # for i in pgrep rcu[^c] ; do taskset -pc 0 $i ; done
  3. Optional: Use the isolcpus parameter on the kernel command line to isolate certain cores from user-space tasks.
  4. Optional: Set the CPU affinity for the kernel’s write-back bdi-flush threads to the housekeeping core: 
                echo 1 > /sys/bus/workqueue/devices/writeback/cpumask

Results

  • Once the system is rebooted, verify if dynticks are enabled: 
    # journalctl -xe | grep dynticks
     
    Mar 15 18:34:54 rhel-server kernel: NO_HZ: Full dynticks CPUs: 1-15.
  • Verify that the dynamic tickless configuration is working correctly: 
    # perf stat -C 1 -e irq_vectors:local_timer_entry taskset -c 1 sleep 3

    This command measures ticks on CPU 1 while telling CPU 1 to sleep for 3 seconds. The default kernel timer configuration shows around 3100 ticks on a regular CPU:

    # perf stat -C 0 -e irq_vectors:local_timer_entry taskset -c 0 sleep 3
     
     Performance counter stats for 'CPU(s) 0':

             3,107      irq_vectors:local_timer_entry

  3.001342790 seconds time elapsed

    With the dynamic tickless kernel configured, you should see around 4 ticks instead:

    # perf stat -C 1 -e irq_vectors:local_timer_entry taskset -c 1 sleep 3
     
     Performance counter stats for 'CPU(s) 1':

                 4      irq_vectors:local_timer_entry

       3.001544078 seconds time elapsed