Recommended system configuration and tuning

• Kernel configuration (text from "make menuconfig" help)

  Author's note: I'd suggest to reject PREEMPT_NONE, PREEMPT_VOLUNTARY & PREEMPT_RTB out of hand? Which leaves PREEMPT__LL, PREEMPT_RT_FULL & DEADLINE up for discussion?

• Preemption model

• PREEMPT_NONE

This is the traditional Linux preemption model, geared towards throughput. It will still provide good latencies most of the time, but there are no guarantees and occasional longer delays are possible. Select this option if you are building a kernel for a server or scientific/computation system, or if you want to maximize the raw processing power of the kernel, irrespective of scheduling latencies.

• PREEMPT_VOLUNTARY

This option reduces the latency of the kernel by adding more "explicit preemption points" to the kernel code. These new preemption points have been selected to reduce the maximum latency of rescheduling, providing faster application reactions, at the cost of slightly lower throughput. This allows reaction to interactive events by allowing a low priority process to voluntarily preempt itself even if it is in kernel mode executing a system call. This allows applications to run more 'smoothly' even when the system is under load. Select this if you are building a kernel for a desktop system.

• PREEMPT__LL

This option reduces the latency of the kernel by making all kernel code (that is not executing in a critical section) preemptible. This allows reaction to interactive events by permitting a low priority process to be preempted involuntarily even if it is in kernel mode executing a system call and would otherwise not be about to reach a natural preemption point. This allows applications to run more 'smoothly' even when the system is under load, at the cost of slightly lower throughput and a slight runtime overhead to kernel code. Select this if you are building a kernel for a desktop or embedded system with latency requirements in the milliseconds range.

• PREEMPT_RTB

This option is basically the same as (Low-Latency Desktop) but enables changes which are preliminary for the full preemptible RT kernel.

• PREEMPT_RT_FULL

All and everything.

Author's note: Adds preemptible softirqs, sleeping spinlocks, and a more preemptible read-copy-update. In other words, it's faster at scheduling threads than PREEMPT__LL. The nvidia driver is broken for a long time with PREEMPT_RT_FULL and mandates an extra patch for the kernel. The module installation script also needs a patch to install.

• DEADLINE

• Ticker vs dynamic ticks
• High resolution timers

• Kernel tuning
• Userspace configuration

• PAM rights

• rtprio
• memlock
• nofile

• /dev/cpu_dma_latency
• /dev/rtc
• /dev/hpet
• Access to audio devices
• Priority of soundcard irq thread, rtirq/udev-rtirq

• user space tooling

• tuna
• cpupower

• Topics to investigate