Optimizing performance

Monitor processes

Each running program is a process that can run in the foreground, display output on a sonsole or GUI, or run in the background.

Each process gets a process ID (PID), and linux uses the PID to manage how the process uses CPU and memory.

When linux boots, it starts the init process

Core of the system, it runs scripts that start all other processes running on the system

pgrep

Returns the PID for the currently running process that matches the pattern:

pgrep system*
1
213
260
522
524
715
785
...

# find process
pgrep sleep
1675989
1680227

# Send hangup signal
$ sudo kill -SIGHUP 1680227

# verify process is no longer running
$ pgrep sleep
1675989

# view in other terminal
sleep 1000
Hangup

pidof

Returns the PID of the specified process:

pidof systemd
5276 2687

ps

View currently running processes. Accepts different kinds of params:

preceded by dash (Unix-style)
no dash (BSD-style)
double dash (GNU-style)

# ps columns
UID          PID    PPID  C STIME TTY          TIME CMD

UID

user ID that is running the program

PID

process ID

PPID

process ID of the parent process

C

processor utilization over the lifetime of the process

STIME

system time when the process started

TTY

terminal device that started the process

TIME

cumulative CPU time required to run the process

CMD

name of the program that started the process

[]

In the CMD column means that the process is is currently swapped out from physical memory into virtual memory (swap space) on the hard drive.

These processes are sleeping, which usually means that they are waiting for an event. When an event triggers, the kernel sends the process a signal, and it reacts depending on which sleep mode it is in:

interruptible sleep: receives the signal and immediately wakes up
uninterruptible sleep: only wakes up based on an external event (ex: hardware available). It saves other signals and acts on them when it wakes up.

A zombie process is when a process ended but its parent hasn’t acknowledged the termination signal because it is sleeping.

# processes in current user shell
ps
    PID TTY          TIME CMD
 362858 pts/0    00:00:00 bash
1554908 pts/0    00:00:00 ps

# every process running on the system
ps -ef
UID          PID    PPID  C STIME TTY          TIME CMD
root           1       0  0 Apr18 ?        00:00:14 /sbin/init splash
root           2       0  0 Apr18 ?        00:00:00 [kthreadd]
root           3       2  0 Apr18 ?        00:00:00 [rcu_gp]
root           4       2  0 Apr18 ?        00:00:00 [rcu_par_gp]
root           5       2  0 Apr18 ?        00:00:00 [slub_flushwq]
root           6       2  0 Apr18 ?        00:00:00 [netns]
root           8       2  0 Apr18 ?        00:00:00 [kworker/0:0H-events_highpri]
...

top

Displays process information in real time. The top section includes general system information:

top - 09:47:32 up 5 days, 11:54,  1 user,  load average: 0.75, 0.52, 0.48       # general sys info
Tasks: 376 total,   1 running, 375 sleeping,   0 stopped,   0 zombie            # general process info
%Cpu(s):  0.8 us,  1.2 sy,  0.2 ni, 97.8 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st # CPU info 
MiB Mem :  31704.9 total,  10716.0 free,   8790.9 used,  12198.0 buff/cache     # physical memory
MiB Swap:  65536.0 total,  65536.0 free,      0.0 used.  21815.3 avail Mem      # swap memory

    PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND                      
   7972 linuxus+  20   0 6294616   3.8g   3.7g S   7.9  12.2 100:58.48 VBoxHeadless                 
   4565 linuxus+  20   0 6446072 372728 147248 S   4.6   1.1  39:17.97 gnome-shell                  
 652102 linuxus+  20   0 2788404 593496 537340 S   2.0   1.8  18:25.31 VBoxHeadless                 
   5781 linuxus+  20   0  582208  78284  44020 S   1.7   0.2   2:16.89 gnome-terminal-              
    314 root     -51   0       0      0      0 S   1.0   0.0   2:14.13 irq/51-SYNA2393              
   1702 mysql     20   0 2376580 399892  36116 S   1.0   1.2   9:49.43 mysqld

CPU details

Symbol	Category	Description
us	User	amount of time the CPU spends running application code
sy	System	amount of time the CPU spends working with sys resources
ni	Nice	amount of time CPU spends running low-priority processes
id	Idle	amount of time CPU is not busy
wa	Waiting	(`iowait`) amount of time CPU spends waiting for disk or network ops to complete
hi	Hardware interrupt	amount of time CPU spends processing hardware interrupts
si	Software interrupt	amount of time CPU spends processing software interrupts

Process columns

PID

process ID

USER

process owner’s username

PR

priority of process

NI

nice value of process

VIRT

amount of virtual memory (swap space) used by the process

RES

amount of physical memory used by the process

SHR

amount of memory that the process is sharing with other processes

S

Process status:

D: interruptible sleep
R: running
S: sleeping
T: traced or stopped
Z: zombie

%CPU

share of CPU time that the process is using

%MEM

share of available physical memory that the process is using

TIME+

total CPU time that the process has used since starting

COMMAND

command-line name of the process

Interactive commands

top sorts processes based on %CPU value by default. There are a lot of options, so here are articles explaining a few:

htop

Improved version of top:

Rocky Linux docs

Manage processes

nice

By default, all processes have the same priority to obtain CPU time and memory resources. nice starts an application with a non-default priority setting:

# lower val = higher priority
nice -n <val> <command>

renice

By default, all processes have the same priority to obtain CPU time and memory resources. renice changes the priority of a process that’s already running:

renice <priority> [-p pids] [-u users] [-g groups]

# example
renice 15 -p 3178

Stopping processes

There are 2 commands that send a process signal to running processes: kill and pkill.

Process signals

Processes communicate with each other using process signals, which are predefined messages that processes recognize and may choose to ignore or act on.

Number	Name	Description
1	`SIGHUP`	Hangs up
2	`SIGINT`	Interrupts
3	`SIGQUIT`	Stops running
9	`SIGKILL`	Unconditionally terminates
11	`SIGSEGV`	Segments violation
15	`SIGTERM`	Terminates if possible
17	`SIGSTOP`	Stops unconditionally but doesn’t terminate
18	`SIGTSTP`	Stops or pauses but continues to run in the bg
19	`SIGCONT`	Resumes execution after STOP or TSTP

kill

Sends a signal process based on PID. By default, sends a SIGTERM signal, but you can specify the signal with the -s option and should send them in the following order:

TERM
SIGINT or SIGHUP
SIGKILL (most forecful)

There is no output, so you have to send use the ps or top command to verify that the command stopped.

kill <PID>

# send SIGTERM
kill 3980

# specify the signal
kill -s SIGHUP 3980

# specify the signal shorthand
kill -SIGHUP 3980

pkill

Stop processes using the name instead of PID. Accepts wildcard characters, too:

sudo pkill http*