Understanding Nice Values

Understanding Nice Values

Nice values are the user-space mechanism for influencing the Linux CPU scheduler. A high nice value means "I am being nice — let others have CPU first." The name comes from the social concept: a nice person steps back and lets others go first. Understanding what nice values actually do — and their limitations — helps you make appropriate decisions about when to use them and when a harder limit (cgroups) is more appropriate.

The nice value scale

Nice value scale:
  -20 ────────────── 0 ────────────── +19
   ↑                  ↑                 ↑
Most aggressive    Default         Most yielding
 (root only)      (everyone)        (everyone)

Example: server with 2 CPU-intensive processes:
  Process A: nice 0  (normal)
  Process B: nice 10 (yields)

  When both are runnable: A gets roughly 10× more CPU time than B
  (Exact ratio depends on kernel version and CFS weights)

Nice vs priority in top and ps

Tools display both the nice value and the calculated scheduling priority. The relationship: priority = 20 + nice for normal processes.

# In ps, use -o to see both:
ps -eo pid,ni,pri,stat,comm --sort=ni | head -10

ps output with nice (NI) and priority (PRI)

  PID  NI PRI STAT COMMAND
 1234 -10   9 Ss   sshd           ← NI=-10, PRI=9, high priority
 5678   0  19 Ss   nginx          ← NI=0, PRI=20, normal
12345  10  29 SN   rsync          ← NI=10, PRI=30, below normal
67890  19  39 S N  gzip           ← NI=19, PRI=39, lowest

# In top: NI and PR columns
# Look for processes with negative NI (high priority) or high NI (low priority)

Real-world impact of nice

# Demonstrate nice impact: run two CPU-intensive tasks
# Without nice (compete equally):
sha256sum /dev/urandom &    # Background CPU-intensive task
sha256sum /dev/urandom &    # Another equally important task
# Both get 50% CPU each

# With nice (background job yields):
nice -n 19 sha256sum /dev/urandom &    # Batch job
sha256sum /dev/urandom &                # Interactive job
# Interactive job gets ~90%+ CPU when runnable, batch job gets remainder

📝 NOTE: Nice values only matter when the CPU is busy. If the CPU is idle, a nice +19 process gets 100% CPU just like any other process — there is nothing to yield to. Nice helps when you have multiple processes competing for limited CPU time.

ionice: I/O priority

The ionice command sets I/O scheduling priority, analogous to nice for CPU. Use it with nice for background jobs that do both CPU and disk work.

# I/O scheduling classes:
# 1 = real-time (bypasses others, can starve system)
# 2 = best-effort (default, like normal priority — level 0-7)
# 3 = idle (only gets I/O when no other process wants it — best for backups)

# Run rsync with idle I/O priority and low CPU priority
ionice -c 3 nice -n 19 rsync -av /home/ /backup/home/

# Set I/O priority of a running process
sudo ionice -c 3 -p PID

# Check current I/O priority
ionice -p PID

cgroup CPU limits as an alternative

Nice values are hints to the scheduler but do not provide hard limits. A nice -n 19 process can still consume 100% CPU if it is the only runnable process. For hard enforcement, use cgroups.

# Limit a process to 25% CPU using cgroups (systemd-run)
# Run a backup with a 25% CPU limit
systemd-run --scope -p CPUQuota=25% nice -n 19 rsync -av /home/ /backup/

# Or apply to an existing process via systemd
sudo systemctl set-property --runtime user-1001.slice CPUQuota=50%

# View current cgroup CPU limits
systemd-cgls
cat /sys/fs/cgroup/system.slice/cpu.max

Conclusion

Nice values control CPU scheduler preference but do not enforce hard limits. Use nice -n 19 to start background jobs that should only use CPU when nothing else needs it. Combine with ionice -c 3 for I/O-heavy background work. For hard CPU limits that prevent a process from ever taking more than a specified percentage, use systemd’s CPUQuota= property or cgroups directly. Check nice values in ps -o ni or the NI column in top/htop.