Skip to content

README.md

Latest commit

 

History

History
287 lines (216 loc) · 7.32 KB

README.md

File metadata and controls

287 lines (216 loc) · 7.32 KB

⚡ Chapter 23: Performance Tuning

Advanced Chapter 23

📑 Table of Contents

Performance Analysis Methodology

USE Method — for every resource, check:

  • Utilization — How busy is it?
  • Saturation — Are there queues? Is work waiting?
  • Errors — Are there errors?
Resource Utilization Saturation Errors
CPU top, mpstat vmstat r > cores dmesg, perf
Memory free, vmstat vmstat si/so, OOM kills dmesg
Disk iostat %util iostat await smartctl, dmesg
Network sar -n DEV ss -s, drops ip -s link, dmesg

CPU Tuning

CPU Governors

# View current governor
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor

# List available governors
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors
# conservative ondemand userspace powersave performance schedutil

# Set governor
echo "performance" | sudo tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

# Permanent with cpupower
sudo apt install linux-tools-common
sudo cpupower frequency-set -g performance
sudo cpupower frequency-info
Governor Behavior
performance Always max frequency (best throughput)
powersave Always min frequency (best battery)
ondemand Scale up quickly on load
schedutil Kernel scheduler decides (default)

CPU Affinity

# Pin process to specific CPUs
taskset -c 0,1 ./myprocess           # CPUs 0 and 1
taskset -cp 0,1 1234                  # Change running process
taskset -p 1234                       # View current affinity

# Using cgroups
sudo cgcreate -g cpu:myapp
sudo cgset -r cpuset.cpus=0-3 myapp
sudo cgexec -g cpu:myapp ./myprocess

Memory Tuning

Swappiness

# Lower = prefer RAM, higher = prefer swap
cat /proc/sys/vm/swappiness           # Default: 60
sudo sysctl vm.swappiness=10          # Set for SSD servers

# vm.vfs_cache_pressure
sudo sysctl vm.vfs_cache_pressure=50  # Default: 100
# Lower = keep dentries/inodes in cache longer

Hugepages

# For databases and large-memory applications
cat /proc/meminfo | grep Huge

# Allocate 1024 hugepages (2MB each = 2GB)
echo 1024 | sudo tee /proc/sys/vm/nr_hugepages
# Permanent
echo "vm.nr_hugepages = 1024" | sudo tee -a /etc/sysctl.conf

# Transparent Hugepages (THP)
cat /sys/kernel/mm/transparent_hugepage/enabled
# Options: [always] madvise never
echo "madvise" | sudo tee /sys/kernel/mm/transparent_hugepage/enabled

OOM Killer Tuning

# View OOM score (higher = more likely to be killed)
cat /proc/$(pidof mysql)/oom_score
cat /proc/$(pidof mysql)/oom_score_adj

# Protect a process from OOM killer
echo -1000 | sudo tee /proc/$(pidof mysql)/oom_score_adj

I/O Tuning

I/O Schedulers

# View current scheduler
cat /sys/block/sda/queue/scheduler
# [mq-deadline] kyber bfq none

# Change scheduler
echo "kyber" | sudo tee /sys/block/sda/queue/scheduler
Scheduler Best For
mq-deadline General purpose, databases (default)
kyber Fast NVMe SSDs
bfq Desktop interactive, USB drives
none NVMe with hardware queues

Read-Ahead

# View/set read-ahead (KB)
blockdev --getra /dev/sda             # Current value
sudo blockdev --setra 2048 /dev/sda   # Set to 2048 sectors

Filesystem Mount Options

# /etc/fstab optimizations
UUID=xxx /     ext4  noatime,nodiratime,discard  0 1
#  noatime     = don't update access times (big performance gain)
#  nodiratime  = don't update directory access times
#  discard     = enable TRIM for SSD (or use fstrim.timer)

# Enable periodic TRIM
sudo systemctl enable fstrim.timer

Network Tuning

# Key sysctl network parameters
sudo sysctl -w net.core.somaxconn=65535
sudo sysctl -w net.ipv4.tcp_max_syn_backlog=65535
sudo sysctl -w net.core.rmem_max=16777216
sudo sysctl -w net.core.wmem_max=16777216
sudo sysctl -w net.ipv4.tcp_rmem="4096 87380 16777216"
sudo sysctl -w net.ipv4.tcp_wmem="4096 87380 16777216"
sudo sysctl -w net.ipv4.tcp_fin_timeout=15
sudo sysctl -w net.ipv4.tcp_tw_reuse=1
sudo sysctl -w net.core.netdev_max_backlog=5000

cgroups v2 — Resource Limits

# Check if cgroups v2 is enabled
mount | grep cgroup2

# Create a cgroup
sudo mkdir /sys/fs/cgroup/mygroup

# Set limits
echo "500000 1000000" | sudo tee /sys/fs/cgroup/mygroup/cpu.max  # 50% CPU
echo "512M" | sudo tee /sys/fs/cgroup/mygroup/memory.max         # 512MB RAM
echo "100" | sudo tee /sys/fs/cgroup/mygroup/io.weight            # I/O weight

# Add process to cgroup
echo $PID | sudo tee /sys/fs/cgroup/mygroup/cgroup.procs

# With systemd (easier)
sudo systemctl set-property myservice CPUQuota=50%
sudo systemctl set-property myservice MemoryMax=512M

tuned — Automatic Profiles

sudo apt install tuned
sudo systemctl enable --now tuned

# List profiles
tuned-adm list

# Apply profile
sudo tuned-adm profile throughput-performance

# Recommended
tuned-adm recommend
Profile Use Case
balanced Default
throughput-performance High throughput servers
latency-performance Low latency
network-latency Network-sensitive apps
virtual-guest Running as VM
desktop Desktop responsiveness
powersave Laptop battery

Kernel Parameters for Performance

# /etc/sysctl.d/99-performance.conf
vm.swappiness = 10
vm.vfs_cache_pressure = 50
vm.dirty_ratio = 20
vm.dirty_background_ratio = 5
net.core.somaxconn = 65535
net.ipv4.tcp_max_syn_backlog = 65535
fs.file-max = 2097152
fs.inotify.max_user_watches = 524288

# Reload
sudo sysctl --system

File Descriptor Limits

# View current limits
ulimit -a
ulimit -n                            # Open files limit

# Increase permanently
sudo vim /etc/security/limits.conf
# *    soft    nofile    65536
# *    hard    nofile    65536

🏋️ Practice Exercises

  1. CPU: Change your CPU governor to performance and back
  2. Memory: Tune swappiness to 10 and measure the effect
  3. I/O: Check and change your disk scheduler
  4. cgroups: Create a cgroup that limits a process to 50% CPU
  5. tuned: Install tuned and apply throughput-performance profile
  6. Limits: Increase the file descriptor limit to 65536
  7. noatime: Add noatime to a mount and benchmark reads
  8. Monitor: Use the USE method to analyze your system resources

← Previous: SELinux & AppArmor · 🏠 Home · Next: LVM & RAID →

⚡