ODROID-XU4 tune network and USB speed

It’s time for another small blog about the ODROID-XU4.
This is just a quick tip to improve your network and USB performance even more. It will optimize your hardware interrupts (IRQ) affinity on your ODROID-XU4.
This guide is for the 3.10.y kernel and debian 8. For other kernel versions the interrupts may have different numbers.

Description

Whenever a piece of hardware, such as disk controller or ethernet card, needs attention from the CPU, it throws an interrupt. The interrupt tells the CPU that something has happened and that the CPU should drop what it’s doing to handle the event. In order to prevent multiple devices from sending the same interrupts, the IRQ system was established where each device in a computer system is assigned its own special IRQ so that its interrupts are unique.
Starting with the 2.4 kernel, Linux has gained the ability to assign certain IRQs to specific processors (or groups of processors). This is known as SMP IRQ affinity, and it allows you control how your system will respond to various hardware events. It allows you to restrict or repartition the workload that you server must do so that it can more efficiently do it’s job.
Source

It’s always a good idea to spread your interrupts evenly across all CPUs. In my case I want to achieve the best performance possible. Therefore I want to use the faster A15 CPU cluster for all important interrupt handling.

There are basically 3 different interrupts on a headless ODROID-XU4 server you should take into consideration:

  • the USB2 port
  • the first USB3 port
  • the second USB3 port (the 1 Gigabit ethernet adapter is connected to this one)

Per default all 3 interrupts for these devices are handled by CPU0, which is a A7 core as you can see in the output below:

IRQ Tuning

First of all make sure that automatic IRQ balancing is disabled:

For debian add the following to your /etc/rc.local  file to pin the interrupt handling to A15 cores 4-6 (CPU4-6):

After a reboot and some file transfer you should see something like this:

Note the numbers for CPU4 and CPU5. CPU0 handled some initial interrupts during the boot, because rc.local isn’t executed immediately.

Benchmarks

Tuning without measuring performance before and afterwards is useless. So, here are some iperf results:

Up to 100 Mbit/s faster. Not bad for such an easy fix 🙂

Read my post in the ODROID forum to get some more information and tuning tips.