Overview

This section summarizes the information about the oeAware plugins and instances, and describes how to use the tuning instances. The collection and sensing plugins are used to collect and integrate data and provide the data for the tuning plugins. When the tuning plugins are used, the collection and awareness instances on which they depend are automatically started. The following mainly describes how to use the tuning plugins.

Collection Plugins

Overview

Collection Plugin	Instance Name	Description	2203-LTS-SP4	2403-LTS-SP1
libpmu.so	pmu_counting_collector	Collect the system PMU performance counters (core).	AArch64	AArch64
libpmu.so	pmu_uncore_collector	Collect the system PMU performance counters (uncore).	AArch64	AArch64
libpmu.so	pmu_sampling_collector	Collect system PMU-related behavior records.	AArch64	AArch64
libpmu.so	pmu_spe_collector	Collect system SPE records.	AArch64	AArch64
libdocker_collector.so	docker_cpu_collector	Collect Docker container information in the current environment.	Supported	Supported
libsystem_collector.so	thread_collector	Collect thread information in the current environment.	Supported	Supported
libsystem_collector.so	kernel_config	Collect or configure kernel parameters.	Supported	Supported
libsystem_collector.so	command_collector	Use sysstat-related collection commands to collect system information.	Supported	Supported

Sensing Plugins

Overview

Sensing Plugin	Instance Name	Description	2203-LTS-SP4	2403-LTS-SP1
libthread_scenario.so	thread_scenario	Obtain key thread information in the current environment.	Supported	Supported
libscenario_numa.so (external plugin)	scenario_numa	Obtain the cross-NUMA memory access ratio in the current environment.	AArch64	AArch64
libanalysis_aware.so	analysis_aware	Analyze service characteristics in the current environment and provide tuning suggestions.	AArch64	AArch64

Tuning Plugins

Overview

Tuning Plugin	Instance Name	Description	2203-LTS-SP4	2403-LTS-SP1
libsystem_tune.so	stealtask_tune	Optimize CPU scheduling, reduce CPU idling, and improve CPU utilization.	AArch64	AArch64
libsystem_tune.so	smc_tune	Improve the network throughput and reduce the latency based on shared memory communication in the kernel space.	Not supported	Supported
libsystem_tune.so	xcall_tune	Bypass non-essential code paths to minimize system call processing overhead.	Not supported	AArch64
libsystem_tune.so	seep_tune	Enable dynamic frequency scaling to reduce overall system power consumption.	AArch64	AArch64
libub_tune.so	unixbench_tune	Optimize the UnixBench test.	Supported	Supported
libdocker_tune.so	docker_cpu_burst	CPUBurst can temporarily provide additional CPU resources for containers to alleviate performance bottlenecks caused by CPU limits when burst loads occur. This ensures and improves the service quality of applications (especially latency-sensitive applications).	AArch64	AArch64
libtune_numa.so (external plugin)	tune_numa_mem_access	Periodically migrate threads and memory to reduce cross-NUMA memory access.	AArch64	AArch64
Gazelle	Not integrated into oeAware	The high-performance user-mode protocol stack greatly improves the network I/O throughput of applications and focuses on database network performance acceleration.	-	-
libdfot.so	dfot_tuner_sysboost	Dynamic feedback optimization (optimization at startup and runtime). Currently, the optimization at startup is implemented.	AArch64	Not supported

Tuning Instance Usage

tune_numa_mem_access

Application Scenarios

Frequent memory access
Performance gains from manual core or NUMA affinity setting

Prerequisites

Operating Environment

AArch64
Physical machine
openEuler kernel (4.19, 5.10, 6.6)

Enabling SPE

This plugin relies on the BIOS SPE feature. Before running the plugin, you need to enable the SPE. Run perf list | grep arm_spe to check whether the SPE is enabled. If it is enabled, the following information is displayed:

shell

arm_spe_0//                                      [Kernel PMU event]

If not, perform the following steps to enable it:

Go to MISC Config --> SPE in the BIOS. If the SPE is set to Disable, switch it to Enable. If you cannot find this option, the BIOS version may be outdated.

Access vim /etc/grub2-efi.cfg of the system, locate the startup item corresponding to the kernel version, and add kpti=off to the end of the startup item. Example:

shell

        linux   /vmlinuz-4.19.90-2003.4.0.0036.oe1.aarch64 root=/dev/mapper/openeuler-root ro rd.lvm.lv=openeuler/root rd.lvm.lv=openeuler/swap video=VGA-1:640x480-32@60me rhgb quiet  smmu.bypassdev=0x1000:0x17 smmu.bypassdev=0x1000:0x15 crashkernel=1024M,high video=efifb:off video=VGA-1:640x480-32@60me kpti=off

Press Esc, enter :wq, and press Enter to save the change and exit.
Run the reboot command to restart the server.

Installing the Plugin

This plugin is an external plugin and is not installed together with oeAware. You need to install it separately.

Check whether the plugin is installed.

shell

oeawarectl -q | grep tune_numa_mem_access

If the plugin does not exist, install it as follows:
Installation method 1: oeawarectl -i numafast
Installation method 2: Select the RPM package of numafast corresponding to the current system kernel version from OEPKGS and manually install it. You are advised to use the latest version of numafast.
After installation, load the plugin by running oeawarectl -l tune_numa_mem_access.

How to Use

Enable this instance.
shell
```
oeawarectl -e tune_numa_mem_access
```
Stop this instance.
shell
```
oeawarectl -d tune_numa_mem_access
```

docker_cpu_burst

Application Scenarios

Common containers, not applicable to K8s
High service load in the container

Prerequisites

Operating Environment
- openEuler kernel (5.10, 6.6)

How to Use

Enable this instance.
shell
```
oeawarectl -e docker_cpu_burst
```
Stop this instance.
shell
```
oeawarectl -d docker_cpu_burst
```

unixbench_tune

Application Scenarios

UnixBench test

How to Use

Enable this instance.
shell
```
oeawarectl -e unixbench_tune
```
Stop this instance.
shell
```
oeawarectl -d unixbench_tune
```

stealtask_tune

Application Scenarios

High service load

Prerequisites

Operating Environment
- openEuler kernel (5.10, 6.6)

How to Use

Enable this instance.
shell
```
oeawarectl -e stealtask_tune
```
Stop this instance.
shell
```
oeawarectl -d stealtask_tune
```

xcall_tune

Application Scenarios

System call overhead reduction

Prerequisites

To be added.

How to Use

shell

oeawarectl -e xcall_tune

seep_tune

Application Scenarios

Energy saving

Prerequisites

AArch64 physical machine
XXX enabled on the BIOS (to be added)

How to Use

Enable this instance.
shell
```
oeawarectl -e seep_tune
```
Stop this instance.
shell
```
oeawarectl -d seep_tune
```

smc_tune

Application Scenarios

Local network communication

Prerequisites

openEuler kernel (6.6)

How to Use

Enable this instance.
shell
```
oeawarectl -e smc_tune
```
Stop this instance.
shell
```
oeawarectl -d smc_tune
```

dfot_tuner_sysboost

Prerequisites

Operating Environment

openEuler kernel (5.10)

Installing the Plugin

This plugin is an external plugin and is not installed together with oeAware. You need to install it separately.

Check whether the plugin is installed.

shell

oeawarectl -q | grep dfot_tuner_sysboost

If the plugin does not exist, install it.
shell
```
yum install D-FOT
```
Load the plugin after installation.
shell
```
oeawarectl -l tune_numa_mem_access
```

How to Use

Enable this instance.
shell
```
oeawarectl -e dfot_tuner_sysboost
```
Stop this instance.
shell
```
oeawarectl -d dfot_tuner_sysboost
```

Gazelle

Application Scenarios

Service performance affected by the network
Low latency and high throughput

How to Use

This tuning capability has not been integrated into oeAware. For details, see Gazelle User Guide.

Overview ​

Collection Plugins ​

Overview ​

Sensing Plugins ​

Overview ​

Tuning Plugins ​

Overview ​

Tuning Instance Usage ​

tune_numa_mem_access ​

Application Scenarios ​

Prerequisites ​

Operating Environment ​

Enabling SPE ​

Installing the Plugin ​

How to Use ​

docker_cpu_burst ​

Application Scenarios ​

Prerequisites ​

How to Use ​

unixbench_tune ​

Application Scenarios ​

How to Use ​

stealtask_tune ​

Application Scenarios ​

Prerequisites ​

How to Use ​

xcall_tune ​

Application Scenarios ​

Prerequisites ​

How to Use ​

seep_tune ​

Application Scenarios ​

Prerequisites ​

How to Use ​

smc_tune ​

Application Scenarios ​

Prerequisites ​

How to Use ​

dfot_tuner_sysboost ​

Prerequisites ​

Operating Environment ​

Installing the Plugin ​

How to Use ​

Gazelle ​

Application Scenarios ​

How to Use ​

Overview

Collection Plugins

Overview

Sensing Plugins

Overview

Tuning Plugins

Overview

Tuning Instance Usage

tune_numa_mem_access

Application Scenarios

Prerequisites

Operating Environment

Enabling SPE

Installing the Plugin

How to Use

docker_cpu_burst

Application Scenarios

Prerequisites

How to Use

unixbench_tune

Application Scenarios

How to Use

stealtask_tune

Application Scenarios

Prerequisites

How to Use

xcall_tune

Application Scenarios

Prerequisites

How to Use

seep_tune

Application Scenarios

Prerequisites

How to Use

smc_tune

Application Scenarios

Prerequisites

How to Use

dfot_tuner_sysboost

Prerequisites

Operating Environment

Installing the Plugin

How to Use

Gazelle

Application Scenarios

How to Use