FMEA for Edge Compute Offloading: Protocol Conformance and Production SLA Failure Modes

Moving inspection, analytics or control logic from a machine to an edge or MEC node (ETSI GS MEC 003) adds new failure modes: latency budget breaches, protocol non-conformance and SLA drift between lab and production. FMEA Tool lets you put these risks through the 7-step AIAG & VDA FMEA Handbook (1st ed., 2019) method, consistent with IEC 60812:2018. Each offloaded function becomes a measurable requirement, and Action Priority (AP) decides where mitigation must happen before rollout. Transport and protocol assumptions should point to MQTT Version 5.0 (OASIS, 2019), OPC UA (IEC 62541), QUIC (RFC 9000) or 3GPP TS 23.501, not to vendor datasheets.

Offload decision criteria

Offload a function only when its end-to-end budget (sensor → decision → actuator) can be written as a percentile requirement, for example p99 plus jitter and loss limits, and when a defined fallback exists for a missed deadline. If a failure effect is rated Severity 9–10 or touches a safety function under ISO 26262:2018 or IEC 61508:2010, keep a deterministic local path. Use the edge only for non-safety analytics or advisory results, and record that boundary as a Step 3 function requirement.

Pre-implementation checklist and recovery

Before you specify the architecture, gather the p99 latency budget, the jitter tolerance, the packet loss and link-outage profile, the MQTT 5.0 QoS level and Session Expiry Interval, the OPC UA security and transport profiles, and the clock sync source. Catalog failure modes such as silent QoS downgrade, message loss when a session expires, broker failover reordering and timestamp drift. Pair each with prevention controls (store-and-forward buffers, idempotent consumers, a local watchdog that fails safe) and detection controls (conformance test suites, contract tests, production p99 telemetry alarms). Roll out one cell or line at a time, with a documented way to revert to local processing.

Illustrative scenario (analytical, not measured)

A vision inspection station sends inference to an on-site MEC node with an assumed 50 ms p99 budget before the part leaves the station. Failure mode: "result arrives after the part has left", so a defective part escapes downstream. Illustrative ratings are S=8, O=6 and D=7, which falls in the High AP band of the AIAG & VDA table. The justified action is a local timer: no verdict within the budget means the part is diverted to a reject or quarantine lane. That turns a latency failure into a safe, detectable outcome. Then rate O and D again from production evidence, not from vendor benchmarks.

How do I check that an edge-offloading FMEA in fmeatool reflects production SLAs and not lab benchmarks?

Write each SLA as a function requirement with a percentile, a measurement window and a measurement point (for example p99 end-to-end over a shift, measured at the actuator). Accept as detection controls only methods that observe the real system: production telemetry, conformance tests run against the deployed broker or server, and fault-injection drills. Treat lab or vendor numbers as assumptions, not evidence for Occurrence. When production p99, loss rates or protocol versions change, reopen the affected failure chains, re-rate S/O/D, check the resulting Action Priority again, and export the updated Excel worksheet as the audit trail for the change.

Back to FMEA Tool · Domains

← Back to FMEA Tool
Engineering & Architecture Brief

FMEA for Edge Compute Offloading: Protocol Conformance and Production SLA Failure Modes

Moving inspection, analytics or control logic from a machine to an edge or MEC node (ETSI GS MEC 003) adds new failure modes: latency budget breaches, protocol non-conformance and SLA drift between lab and production. FMEA Tool lets you put these risks through the 7-step AIAG & VDA FMEA Handbook (1st ed., 2019) method, consistent with IEC 60812:2018. Each offloaded function becomes a measurable requirement, and Action Priority (AP) decides where mitigation must happen before rollout. Transport and protocol assumptions should point to MQTT Version 5.0 (OASIS, 2019), OPC UA (IEC 62541), QUIC (RFC 9000) or 3GPP TS 23.501, not to vendor datasheets.

Offload decision criteria

Offload a function only when its end-to-end budget (sensor → decision → actuator) can be written as a percentile requirement, for example p99 plus jitter and loss limits, and when a defined fallback exists for a missed deadline. If a failure effect is rated Severity 9–10 or touches a safety function under ISO 26262:2018 or IEC 61508:2010, keep a deterministic local path. Use the edge only for non-safety analytics or advisory results, and record that boundary as a Step 3 function requirement.

Pre-implementation checklist and recovery

Before you specify the architecture, gather the p99 latency budget, the jitter tolerance, the packet loss and link-outage profile, the MQTT 5.0 QoS level and Session Expiry Interval, the OPC UA security and transport profiles, and the clock sync source. Catalog failure modes such as silent QoS downgrade, message loss when a session expires, broker failover reordering and timestamp drift. Pair each with prevention controls (store-and-forward buffers, idempotent consumers, a local watchdog that fails safe) and detection controls (conformance test suites, contract tests, production p99 telemetry alarms). Roll out one cell or line at a time, with a documented way to revert to local processing.

Illustrative scenario (analytical, not measured)

A vision inspection station sends inference to an on-site MEC node with an assumed 50 ms p99 budget before the part leaves the station. Failure mode: "result arrives after the part has left", so a defective part escapes downstream. Illustrative ratings are S=8, O=6 and D=7, which falls in the High AP band of the AIAG & VDA table. The justified action is a local timer: no verdict within the budget means the part is diverted to a reject or quarantine lane. That turns a latency failure into a safe, detectable outcome. Then rate O and D again from production evidence, not from vendor benchmarks.

Technical FAQ: How do I check that an edge-offloading FMEA in fmeatool reflects production SLAs and not lab benchmarks?

Write each SLA as a function requirement with a percentile, a measurement window and a measurement point (for example p99 end-to-end over a shift, measured at the actuator). Accept as detection controls only methods that observe the real system: production telemetry, conformance tests run against the deployed broker or server, and fault-injection drills. Treat lab or vendor numbers as assumptions, not evidence for Occurrence. When production p99, loss rates or protocol versions change, reopen the affected failure chains, re-rate S/O/D, check the resulting Action Priority again, and export the updated Excel worksheet as the audit trail for the change.

Related Architecture Specifications (FMEATOOL): fmeatools.com · freefmea.com

More Technical Articles on fmeatool.com

Zero-Trust Cryptography and FMEA for Fault-Isolated Failover →Analyzing Timing Failure Modes with FMEA: Deterministic Telemetry, Sub-Millisecond Latency Budgets and HIL/Digital-Twin Verification →FMEA Reference Architecture: Aligning DFMEA/PFMEA Data with AIAG-VDA, SAE J1739 and IEC 60812 for Interoperable Exchange →Using FMEA for Autonomous Diagnostics, Predictive Maintenance and Cross-Layer Security: A Reference Architecture →DFMEA for Edge Compute Offloading: Deadlines, Protocol Conformance and SLA Evidence →