Controlled Degradation, Not Dispatch Reordering: Semantics-Preserving Overload Control in Lingua Franca
Year: 2026
Type: article
Abstract: Logical-time reactive languages such as Lingua Franca (LF) make timing part of the programming model, but overload handling is usually left to implementation-level scheduling. This letter formulatescontrolled degradation, which sheds application-declareddegradablereactions while preserving LF determinism, and asks a focused question: when high-criticality (HC) and low-criticality (LC) reactions are released at the same logical-time tag, can the runtime shed optional LC work, without violating LF semantics, to protect HC timing and bound overload propagation across tags? We give the mechanism as a pressure-driven, per-tag LC budget enforced at the ready-set boundary with semantic guards and safe fallback, argue that shedding a degradable reaction is observationally equivalent to its non-triggering, and show that, once retained per-tag work exceeds capacity, no work-conserving dispatch order can prevent HC misses; only shedding can. On a multi-worker LF/C runtime on a PREEMPT_RT Raspberry Pi at a natural 2 ms period, degradation keeps HC within its deadline (HC-miss falls from 100% to ≈ 0% at the overload onset), bounds the longest run of overrun tags to at most tens (vs. the full trace), and exposes a tunable LC/timeliness trade-off, with LF determinism preserved by construction; the benefit depends on worker parallelism, which we characterize.
Source: IEEE Embedded Systems Letters
Institutions: Nagoya University, Denso (Japan)
Language: English
Topic: Real-Time Systems Scheduling
Subfield: Hardware and Architecture
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $2,645
OpenAlex ID: W7212112488