We’re hosting the Real-Time Linux User Forum on Tuesday, October 6, 13:30–17:00, the day before Open Source Summit Europe in Prague.
The User Forum is intended to bring together PREEMPT_RT developers and tooling experts with users who are building products and solutions with PREEMPT_RT. It is an opportunity for users to share real-world experiences, and exchange ideas and collaborate with developers on the challenges and opportunities shaping Real-Time Linux.
1:30-2:00 PM: PREEMPT_RT configuration: Best practices with kconfig_checker
Speaker: Ahmed S. Darwish, Linutronix GmbH
A PREEMPT_RT kernel must avoid many subtle configuration pitfalls to meet its intended latency requirements. This session has two parts: first, a summary of PREEMPT_RT kernel configuration best practices, drawn from kernel development experience and real-world deployments; second, a demonstration of “kconfig_checker”, a tool intended for Linux upstream submission that helps identify problematic kernel options across multiple configuration profiles, including PREEMPT_RT.
2:00-2:30 PM: cpusets and other recent updates in tuna and rteval
Speaker: John Kacur, Red Hat
Recent updates in tuna and rteval with a focus on cpusets for creating environments to get the best realtime performance. Cpusets in rteval can work with or independently from isolcpus. Other topics could include the new housekeeping flags, core sharing validation and more.
2:30-3:00 PM: Let the Kernel Monitor Your Real-Time Application
Speaker: John Ogness, Linutronix GmbH
The Linux kernel has a new Runtime Verifier (RV) monitor that reacts if any real-time application performs an operation that could lead to unexpected or unbounded latencies. This includes evaluating sleep and wake events in order to identify likely priority inversion scenarios.
This presentation briefly mentions the necessary precautions all Linux real-time developers must consider. Then it explains how the rtapp monitor is enabled and used. And finally, several examples will be shown of the rtapp monitor identifying issues in real-time applications as well as pointing out how such issues could be fixed.
The new rtapp RV monitor is easy to use and truly belongs in the toolbox of anyone developing, testing, or verifying Linux real-time systems.
3:00-3:30 PM Break
3:30-4:00 PM: Finding sources of Latency via Tracing
Speaker: Steven Rostedt, Qube-RT
A real-time system has worse-case latency requirements. When an unacceptable latency is discovered, it is best to find where that happened. The tracing subsystem of the Linux kernel is a strong tool to investigate those latencies. This talk will go over some old and new methods of using tracing to find the sources of latency that need to be fixed.
4:00-4:30 PM: Beyond Cyclictest: Deploying PREEMPT_RT in Mixed-Criticality Industrial Systems
Speaker: Jitendra sharma, Qualcomm India Private Limited
Industrial edge systems increasingly need to consolidate real-time control, industrial networking, AI inference, HMI applications, and cloud connectivity onto a single hardware platform. While PREEMPT_RT provides the foundation for deterministic Linux systems, achieving predictable behavior in mixed-criticality environments introduces challenges that are not visible in synthetic benchmarks.
This session presents our experience deploying and evaluating PREEMPT_RT on modern ARM64 industrial computing platforms. We discuss how real-time workloads coexist with non-real-time applications, and how system design decisions impact temporal isolation and worst-case latency.
Topics include CPU isolation, IRQ steering, shared-resource contention, memory and cache interference, and methodologies for characterizing real-time performance under system stress.
Attendees will gain practical insights into building predictable industrial Linux systems through an end-to-end design approach, with PREEMPT_RT as a key building block.
4:30-5:00 PM: Where PREEMPT_RT meats the dynamics of container runtimes
Speaker: Florian Bezdeka, Siemens
In the past RT enabled systems were kind of static. Well chosen hardware with a well understood RT workload on top, both parts – hardware and software – under full control. We call that the good old (and boring) times here 😉
We are facing a different situation here. Customers want to consolidate and use their own hardware. At the same time the software part needs a flexible deployment and update mechanism as different parts of the application landscape might be delivered by different entities.
This is where PREEMPT_RT meets the dynamics of container runtimes. We have to prepare the platform for RT workloads that will be running next to a couple of non-RT applications at the same time. As you might imagine, there are a couple of challenges to solve. Resource management is definitely one of them. (RT) applications have to express their needs. From isolated CPU cores to exclusive access to network interfaces: all parts need a proper configuration.
In this talk I will start with explaining our use case by providing some more details. Afterwards we will start a short journey through problems and bugs we were fighting against. Finally I will close with an outlook by describing challenges that we are trying to solve right now.
