The Linux Model Extractor for Real-Time Workloads

LiME is a tool designed to dynamically extract real-time task models from Linux workloads. By observing the temporal behavior of real-time threads, LiME automatically maps their activity to well-established real-time task models, including:

sporadic and periodic tasks
arrival curves (upper and lower)
cumulative execution-time curves ($\mathrm{WCET}(n)$)
self-suspension models (dynamic and segmented)

Unlike traditional approaches, LiME operates fully automatically on unmodified Linux kernels, requiring no prior expertise in real-time theory or Linux kernel internals.

You may find a lot more information about the tool and its theoretical foundations in the LiME paper published at RTAS'25.

Why LiME?

With the PREEMPT_RT patch now part of the mainline Linux kernel, Linux has solidified its role as a major platform for hosting real-time workloads in automotive, aerospace, robotics, and other critical domains. However, despite its widespread use, many Linux-based real-time applications lack formal modeling and analysis, largely due to the absence of automated system introspection tools.

LiME bridges this gap by enabling automated, in situ modeling of real-time tasks, making schedulability analysis more easily accessible to non-experts. We encourage you to try LiME if you are interested in:

validating timing behavior of time-sensitive threads on Linux,
analysing system timing characteristics using schedulability analysis,
debugging unexpected latencies,
continuosly monitoring real-time applications,
learning more about the relationship between the Linux internals and real-time scheduling theory,
substantiating model assumptions made in more theoretical work on the formal analysis of real-time systems,
and whatever use case you can come up with. 🙂

Current Status

🎉 First public version now available! 🎉 Check out LiME on GitHub: https://github.com/LiME-org/lime-rtw

Citation

When using LiME for academic work, please cite the RTAS'25 paper by Brandenburg et al.