Technique How to do it/Commands to do it Does it need sudo access Existing setting Modified setting

1.SCHED_FIFO (First-In, First-Out):

Real-time Scheduling Policy
chrt --fifo --pid <priority> <pid>
(Elevate your process's SCHED_OTHER 1.sched_setscheduler()
YES
scheduling priority to minimize (May be) 2.sched_setattr()
2.SCHED_RR (Round-Robin):
preemptions.)
chrt --rr --pid <priority> <pid>

1. Using systemctl (Systemd):

Stop the service:

sudo systemctl stop <service_name>
Disable the service:
sudo systemctl disable <service_name>
Reduce System Load Mask the service:
(Stop Unnecessary Services & sudo systemctl mask <service_name> YES - _
Daemons)

2. Using chkconfig (SysVinit):

Disable the service:

/sbin/chkconfig <service_name> off
Restore the service:
/sbin/chkconfig <service_name> reset
1. taskset Command
taskset -c 0 ./your_c_program
2.To launch pinned to a list of CPUs
(e.g., CPUs 0 and 1):
CPU Affinity No No CPU Affinity by Default Pinned to Specific Core(s)
taskset -c 0,1 ./your_c_program
3.To find the current affinity of a running
process (e.g., PID 1234):
taskset -p 1234
monitor the /proc/reclaimed_pages file:
watch -n 1 cat /proc/reclaimed_pages
view the existing min_free_kbytes setting:
Indirectly modifying kswapd through /proc/sys/vm/min_free_kbytes and
cat /proc/sys/vm/min_free_kbytes reduction in the frequency and
monitoring /proc/reclaimed_pages can be a technique to reduce YES values of /proc/reclaimed_pages
raising the freemem threshold(generally by 2x): amount of reclaimed pages
aggressive page reclaim & jitter mitigation
sudo sysctl -w vm.min_free_kbytes=new_value
to make change persistent: sudo nano /etc/sysctl.conf
save changes: sudo sysctl -p
Develop a Custom Kernel Module:
1. Specify free memory threshold:
static unsigned long min_free_pages_threshold, reclaim_pages_count;
define static ssize_t min_free_pages_read(), min_free_pages_write();
2. specify the number of pages to reclaim:
define static ssize_t reclaimed_pages_read(), reclaimed_pages_write();
3. monitor disk read requests, amount of free memory & desired reclaim
behavior standard page reclaim behavior is Modified Threshold and page reclaim
Proactive Page Reclaim via a Custom Kernel Module and Thread YES
4. Define kernel thread function: static int reclaim_thread_fn governed by kswapd amount
5. module initialization: Create a Makefile(.ko) using make command
6. Load the Kernel Module : sudo insmod your_module.ko
7. Interact with the /proc Files :
cat /proc/min_free_pages
cat /proc/reclaimed_pages
sudo sh -c "echo 16384 > /proc/min_free_pages
sudo sh -c "echo 1024 > /proc/reclaimed_pages
 1.Check current kernel configuration:
zgrep CONFIG_NO_HZ_FULL /proc/config.gz
2.Use a kernel configuration tool (e.g., make menuconfig, make xconfig)
to enable CONFIG_NO_HZ_FULL under the "Power management and
ACPI options"
or "Processor type and features" menu
(or) Using make menuconfig and enabling CONFIG_NO_HZ_FULL), periodic scheduling ticks enabled on
CPU Isolation using the isolcpus Kernel Parameter. 3. build the kernel:make -j<n> all CPUs.
(programmable timer is configured to generate an interrupt signal at a install modules: sudo make modules_install YES CONFIG_NO_HZ_FULL might or might
specific frequency: configuration parameter, like CONFIG_HZ in Linux) install the kernel : sudo make install) not be enabled in the default kernel
4. Update GRUB configuration
sudo nano /etc/default/grub
5.specify which CPUs should run in "full tickless" mode:
GRUB_CMDLINE_LINUX_DEFAULT="quiet splash nohz_full=2,3"
sudo update-grub
5. sudo reboot
6. verify tickless mode: cat /sys/devices/system/cpu/cpu2/power/tickless==1