Maximizing Wi-Fi Throughput: Fine-Tuning Zephyr for Peak Performance with ESP32 SoCs in IoT Applications

Those who develop IoT applications based on Zephyr OS often need to optimize the communication performance. It can be done by tweaking the Wi-Fi network stack parameters. However, how do you evaluate the resulting changes and find the most suitable configuration? In this article, we will overview a method to optimize communication performance by using the iperf and zperf tools.

iperf is a tool that allows generating network traffic to measure the maximum achievable network bandwidth. zperf is a very similar tool but designed for Zephyr OS.

1. Prepare the testing environment

#

To evaluate the communication performance, we are going to use the following setup:

• Wi-Fi home router
• ESP32-S3-DevKitC-1
• Computer running Ubuntu 22.04

To simplify packet generation and consumption on the Wi-Fi network, we will use iperf on the computer and zperf on the ESP32-S3-DevKitC-1. zperf is included in the standard Zephyr distribution.

This structured testing approach allows to systematically analyze the impact of Zephyr parameter adjustments on Wi-Fi communication in various real-world scenarios, including:

• ESP32 Sending UDP Packets to PC:

  

• PC Sending UDP Packets to ESP32

  

• ESP32 Sending TCP Packets to PC

  

• PC sending TCP Packets to ESP32:

  

2. Configure your computer

#

We are going to install the following on the computer:

• iperf
• Zephyr OS

2.1. Install iperf

#

On a terminal window, execute the following command:

sudo apt-get install iperf

Please note that administrator privileges are required to successfully complete this installation.

2.2. Install Zephyr OS

#

To setup Zephyr OS and its dependencies, follow the instructions in the subsections.

2.2.1. Install Dependencies

#

On a terminal window, execute the following command:

sudo apt install --no-install-recommends \
  git cmake ninja-build gperf ccache dfu-util device-tree-compiler wget \
  python3-dev python3-pip python3-setuptools python3-tk python3-wheel \
  xz-utils file make gcc gcc-multilib g++-multilib libsdl2-dev libmagic1

2.2.2. Install West

#

Install the Python script that manages the Zephyr OS build system by running:

pip install west

2.2.3. Initialize Zephyr

#

Now, initialize Zephyr on your machine by running:

west init ~/zephyrproject
cd ~/zephyrproject
west update

2.2.4. Install Python dependencies

#

After initializing Zephyr, install additional Python dependencies by running:

pip install -r ~/zephyrproject/zephyr/scripts/requirements.txt

2.2.5. Download and Install Zephyr SDK

#

For cross-compiling zperf, download and install the Zephyr SDK by running:

cd ~
wget https://github.com/zephyrproject-rtos/sdk-ng/releases/download/v0.16.4/zephyr-sdk-0.16.4_linux-x86_64.tar.xz
wget -O - https://github.com/zephyrproject-rtos/sdk-ng/releases/download/v0.16.4/sha256.sum | shasum --check --ignore-missing
tar xvf zephyr-sdk-0.16.4_linux-x86_64.tar.xz
cd zephyr-sdk-0.16.4

./setup.sh

2.2.6. Install Espressif binary blobs

#

To successfully build your ESP32-S3 Wi-Fi application on Zephyr, install the hal_espressif binary blobs:

west blobs fetch hal_espressif

2.2.7. Install udev rules

#

Additionally, install udev rules to allow flashing ESP32-S3 as a regular user:

sudo cp /opt/zephyr-sdk-0.16.4/sysroots/x86_64-pokysdk-linux/usr/share/openocd/contrib/60-openocd.rules /etc/udev/rules.d
sudo udevadm control --reload

3. Configure your ESP32-S3-DevKitC-1

#

To run zperf on your ESP32-S3-DevKitC-1, do the following:

• Create the overlay file
• Build and flash zperf

3.1. Create the overlay file

#

• Create the file zephyr/samples/net/zperf/boards/esp32s3_devkitc.overlay and add the following content:

  /*
  * Copyright (c) 2024 Espressif Systems (Shanghai) Co., Ltd.
  *
  * SPDX-License-Identifier: Apache-2.0
  */
  &wifi {
      status = "okay";
  };
  

• To the file zephyr/samples/net/zperf/prj.conf, add the following:

  CONFIG_NET_BUF_DATA_SIZE=1500
  CONFIG_NET_IF_UNICAST_IPV4_ADDR_COUNT=1
  CONFIG_NET_MAX_CONTEXTS=5
  CONFIG_NET_TC_TX_COUNT=1
  CONFIG_NET_SOCKETS=y
  CONFIG_NET_SOCKETS_POSIX_NAMES=y
  CONFIG_NET_SOCKETS_POLL_MAX=4
  CONFIG_POSIX_MAX_FDS=8
  CONFIG_INIT_STACKS=y
  CONFIG_TEST_RANDOM_GENERATOR=y
  CONFIG_NET_L2_ETHERNET=y
  CONFIG_NET_SHELL=y
  CONFIG_NET_L2_WIFI_SHELL=y
  CONFIG_NET_CONFIG_SETTINGS=y
  CONFIG_LOG=y
  CONFIG_SHELL_CMDS_RESIZE=n
  CONFIG_NET_IPV6=n
  CONFIG_NET_DHCPV4=n
  CONFIG_NET_CONFIG_MY_IPV4_ADDR="<STATION IP ADDRESS>"
  CONFIG_NET_CONFIG_MY_IPV4_GW="<GATEWAY IP ADDRESS>"
  CONFIG_NET_CONFIG_MY_IPV4_NETMASK="255.255.255.0"
  CONFIG_NET_TCP_MAX_RECV_WINDOW_SIZE=50000
  

  Replace <STATION IP ADDRESS> and <GATEWAY IP ADDRESS> with the actual IP addresses relevant to your network configuration.

3.2. Build and flash zperf

#

Note: After updating any configuration parameters in zephyr/samples/net/zperf/prj.conf , build zperf and flash it onto the ESP32-S3-DevKitC-1.

Ensure that the ESP32-S3-DevKitC-1 is connected to the computer via USB, then run a testing sequence using the following west commands:

west build -b esp32s3_devkitc zephyr/samples/net/zperf --pristine
west flash
west espressif monitor

4. Run the tests

#

Before starting the testing sequence, always remember to build and flash zperf.

After the build and flash commands, you will access the zperf terminal. Connect the ESP32-S3-DevKitC-1 to the Wi-Fi router using the following steps:

wifi connect <SSID> <PASSWORD>
net ping <PC_IP>

Testing run output:

*** Booting Zephyr OS build zephyr-v3.5.0-2714-g031c842ecb76 ***
[00:00:00.387,000] <inf> net_config: Initializing network
[00:00:00.387,000] <inf> net_config: Waiting interface 1 (0x3fcc8810) to be up...
[00:00:00.388,000] <inf> net_config: Interface 1 (0x3fcc8810) coming up
[00:00:00.388,000] <inf> net_config: IPv4 address: 192.168.15.2
uart:~$ wifi connect <SSID> <PASSWORD>
Connection requested
Connected
uart:~$ net ping 192.168.15.8
PING 192.168.15.8
28 bytes from 192.168.15.8 to 192.168.15.2: icmp_seq=1 ttl=64 time=219 ms
28 bytes from 192.168.15.8 to 192.168.15.2: icmp_seq=2 ttl=64 time=434 ms
28 bytes from 192.168.15.8 to 192.168.15.2: icmp_seq=3 ttl=64 time=356 ms
uart:~$

Now, open a second terminal where you will run iperf.

4.1. ESP32 Sending UDP Packets to PC:

#

On the iperf terminal, type:

iperf -s -l 1K -u -B 192.168.15.8

On the zperf terminal, type:

zperf udp upload 192.168.15.6 5001 10 1K 5M

4.2. PC Sending UDP Packets to ESP32:

#

On the zperf terminal, type:

zperf udp download 5001

On the iperf terminal, type:

iperf -l 1K -u -c 192.168.15.2 -b 10M

4.3. ESP32 Sending TCP Packets to PC

#

On the iperf terminal, type:

iperf -s -l 1K -B 192.168.15.8

On the zperf terminal, type:

zperf tcp upload 192.168.15.8 5001 10 1K 5M

4.4. C Sending TCP Packets to ESP32

#

On the zperf terminal, type:

zperf tcp download 5001

On the iperf terminal, type:

zperf tcp download 5001

5. Results

#

To illustrate the tangible impact of adjusting network-sensitive parameters for ESP32-S3 Wi-Fi throughput, we conducted a series of tests, each time modifying the parameter CONFIG_NET_TCP_MAX_RECV_WINDOW_SIZE. The result highlights the performance progression before and after parameter changes:

5.1. Initial Configuration

#

The parameter CONFIG_NET_TCP_MAX_RECV_WINDOW_SIZE commented out.

5.2. Modified Configuration (Increased Window Size):

#

The parameter CONFIG_NET_TCP_MAX_RECV_WINDOW_SIZE=20000

5.3. Further Modified Configuration (Increased Window Size):

#

The parameter CONFIG_NET_TCP_MAX_RECV_WINDOW_SIZE=50000

5.4. Comparative results chart

#

These results are provided below.

These are valuable insights into the dynamic relationship between network parameters and ESP32-S3 Wi-Fi throughput. It is possible to fine-tune these parameters to achieve optimal performance in diverse network scenarios.

6. Conclusion

#

This article presented how to measure communication performance in response to changes in the values of the Zephyr OS Wi-Fi network stack configuration parameter. The ESP32-S3-DevKitC-1 development board was used.

By installing, configuring, and utilizing the iperf and zperf tools, and following the procedure outlined in this article, it is possible to tangibly observe the performance evolution as the configuration parameters are modified.