Linux - Optimize System Performance with Tuned
Intro
Optimizing system performance is critical for ensuring efficient resource utilization and maintaining reliability under varying workloads. Tuned is a dynamic tuning daemon for Linux that simplifies performance optimization by applying predefined or custom profiles tailored to specific use cases. This guide explores advanced concepts in using Tuned, including custom profile creation, dynamic tuning, and integration with system components like CPU, memory, and I/O schedulers.
Step 1: Installing and Managing Tuned
1.1 Install Tuned
Ensure Tuned is installed on your system:
1
2
3
4
5
# For Debian-based systems
sudo apt install tuned -y
# For Red Hat-based systems
sudo yum install tuned -y
1.2 Enable and Start the Service
Enable and start the Tuned daemon to ensure it runs at boot:
1
2
sudo systemctl enable tuned
sudo systemctl start tuned
1.3 Verify Installation
Check the status of the Tuned service:
1
sudo systemctl status tuned
Step 2: Exploring Predefined Profiles
Tuned provides several predefined profiles optimized for different workloads.
2.1 List Available Profiles
View all available profiles:
1
tuned-adm list
2.2 Apply a Profile
Activate the throughput-performance profile for high throughput workloads:
1
tuned-adm profile throughput-performance
2.3 Verify Active Profile
Check the currently active profile:
1
tuned-adm active
2.4 Common Profiles Overview
balanced: General-purpose optimization balancing performance and energy savings.throughput-performance: Optimized for high I/O throughput.latency-performance: Reduces latency for time-sensitive workloads.virtual-guest: Optimized for virtualized environments.
Step 3: Custom Profile Creation
Custom profiles allow fine-tuning of system parameters to meet specific requirements.
3.1 Create a Custom Profile
Create a directory for your custom profile:
1
sudo mkdir /etc/tuned/custom-profile
Create a configuration file (tuned.conf) for the profile:
1
sudo nano /etc/tuned/custom-profile/tuned.conf
Example Configuration:
1
2
3
4
5
6
7
8
9
10
11
12
[main]
include=throughput-performance
[cpu]
governor=performance
[disk]
readahead=4096
[vm]
transparent_hugepages=never
swappiness=10
This profile:
- Sets the CPU governor to
performance. - Increases disk read-ahead to 4096 KB.
- Disables transparent hugepages and reduces swappiness to optimize memory usage.
3.2 Activate the Custom Profile
Apply the custom profile:
1
tuned-adm profile custom-profile
Step 4: Dynamic Tuning with Plugins
Tuned uses plugins to dynamically adjust system parameters based on workload.
4.1 CPU Tuning
Optimize CPU frequency scaling and governor settings:
1
2
3
4
[cpu]
governor=performance
energy_perf_bias=performance
min_perf_pct=100
This ensures maximum CPU performance by disabling power-saving features.
4.2 Disk I/O Optimization
Optimize disk I/O settings for faster read/write operations:
1
2
3
4
[disk]
readahead=4096
elevator=deadline
write_cache=on
- Sets read-ahead to 4096 KB.
- Uses the
deadlinescheduler for predictable I/O performance. - Enables write caching for faster writes.
4.3 Network Optimization
Tune network parameters for low latency or high throughput:
1
2
3
4
5
6
7
[net]
tcp_rmem_min=4096
tcp_rmem_default=87380
tcp_rmem_max=16777216
tcp_wmem_min=4096
tcp_wmem_default=16384
tcp_wmem_max=16777216
These settings adjust TCP buffer sizes to optimize network throughput.
Step 5: Monitoring and Debugging Tuned
5.1 Analyze Profile Settings
View detailed information about the active profile:
1
2
3
tuned-adm recommend # Suggests an optimal profile based on workload.
tuned-adm list # Lists all available profiles.
tuned-adm active # Displays the currently active profile.
5.2 Debugging Mode
Enable debugging to troubleshoot issues with Tuned:
1
2
sudo tuned -D --debug > tuned-debug.log 2>&1 &
tail -f tuned-debug.log
Step 6: Advanced Use Cases
6.1 Combining Profiles
Combine multiple profiles to leverage their benefits:
1
2
3
4
5
6
7
8
9
[main]
include=throughput-performance,virtual-host
[cpu]
governor=performance
[disk]
elevator=noop
readahead=8192
This combines throughput-performance and virtual-host while overriding specific settings.
6.2 Applying Profiles Temporarily
Apply a profile temporarily without affecting persistent settings:
1
tuned-adm profile latency-performance --runtime-only
The profile will revert after a reboot.
Step 7: Performance Validation
After applying a profile, validate its impact using monitoring tools:
7.1 CPU Usage
Monitor CPU performance:
1
2
3
htop # Interactive process viewer.
lscpu # Displays CPU architecture details.
mpstat # Reports CPU usage statistics.
7.2 Disk Performance
Test disk I/O performance:
1
2
fio --name=test --rw=read --bs=4k --size=1G --numjobs=4 --runtime=60 --group_reporting
iostat -x # Displays detailed disk I/O statistics.
7.3 Network Throughput
Measure network bandwidth:
1
2
3
iperf3 -c <server-ip>
netstat -i # Displays network interface statistics.
tc qdisc show # Shows traffic control configurations.
Conclusion
Tuned simplifies Linux system optimization by dynamically adjusting system parameters based on predefined or custom profiles tailored to specific workloads. By leveraging advanced features like custom profiles, dynamic tuning plugins, and integration with key subsystems (CPU, memory, disk I/O, and networking), you can achieve significant performance improvements while maintaining flexibility and control over your environment.
Experiment with different profiles and validate their effects using monitoring tools to ensure optimal results tailored to your workload needs.