Performance Tuning

Clinical RTOS latency audit

A focused pass on interrupt paths, scheduling assumptions, and worst-case response under clinical workloads.

3 weeks · Remote-first with two on-site days in Budapest · Updated 2026-05-12

Indicative fee: 3 200 000 Ft · informational only

Illustration for Clinical RTOS latency audit

Scope narrative

We profile your real-time operating system configuration against the workloads you describe for patient-facing modes. The audit pairs traceable measurements with a prioritized backlog of changes that reduce tail latency without destabilizing drivers. Deliverables include annotated timelines, a risk note for each proposed change, and a short readout your verification lead can fold into the next cycle.

Included focus areas

  • Trace capture plan aligned to your hazard analysis inputs
  • Worst-case path review for ISRs, DMA, and SPI/I2C stacks
  • Power-aware scheduling checks where sleep states interact with alarms
  • Memory allocator hot spots and fragmentation risk notes
  • Regression checklist for your existing automated test lane
  • Optional pair sessions with your firmware owners

Outcomes you can archive

  • A ranked set of latency reductions with measurement evidence
  • Clearer ownership between drivers, middleware, and application tasks
  • A calmer verification cycle with fewer unexplained timing regressions

FAQ

Do you modify production code during the audit?

We prefer read-only analysis first. If we propose a patch, it lands in a branch you control, with rationale and test notes.

What is not included?

We do not certify the device. We supply engineering evidence your quality standards process can absorb.

Which artifacts should we prepare?

Recent scheduling configuration, representative workload captures, and the hazard analysis excerpts tied to timing.

Experience notes

The RTOS audit named two ISR chains we had been treating as noise. Verification finally stopped reopening the same ticket.
Eszter N. · Director of Systems · NoviSense MedTech · 5/5 · survey