Redis Real-Time Decisioning Demo | Commercial Fleet Predictive Maintenance

Stage 1: The Architecture

Real-time predictive maintenance decisioning for commercial fleet vehicles

Fleet downtime is not a reporting problem — it is a live decisioning problem. OBD-II telematics, Diagnostic Trouble Codes (DTCs), Field Service Actions (FSAs), route assignments, and parts availability must converge before a vehicle leaves the yard on a high-risk run. Redis sits between existing fleet systems and the maintenance decision path, assembling real-time vehicle context for every fault event.

Data Sources

Vehicle Telematics / OBD-II

DTC codes, sensor readings, engine health metrics, EV battery and range state

Fleet Management System

Vehicle assignments, routes, driver schedules, utilization history

DTC + FSA Database

Fault code definitions, field service action history, severity index

Parts & Service Network

Dealer inventory, service bay availability, mobile service capacity

Kafka / Telemetry Stream

Real-time fault events, idle, speed, harsh acceleration, and fuel feeds

Ingest Layer

RDI

Syncs vehicle, maintenance, and fleet assignment state

Redis Feature Form

Serves predictive failure, service urgency, and route risk features online

Unified Context Layer

Redis RAM

Hot fault path, active DTC alerts, live vehicle health state

Redis Flex

Warm maintenance history, failure patterns, and vehicle embeddings

Feature Store

Failure probability, component wear, utilization, and route risk

Redis Context Retriever

Assembles the Vehicle 360 — asset state, maintenance history, and operational context — and exposes it as structured MCP tools for the decision engine

Decision Engine

Fault Severity Rules

DTC threshold scoring, FSA priority, and safety-critical flag logic

Predictive Failure Scorer

ML-based component failure probability from telematics patterns

Service Arbitration

Route risk, parts availability, and schedule constraint resolution

Redis Search

Failure typology matching across historical fleet fault patterns

Output Actions

Proactive Service

Schedule pre-route maintenance before vehicle departs

Route Reassignment

Swap vehicle assignment to available healthy unit

Monitor & Alert

Flag for next scheduled service with active telemetry watch

Emergency Dispatch

Mobile service or roadside assistance with full fault context

Decision Target

<15 ms

Primary Goal

Uptime + fleet utilization

Redis Role

Real-time vehicle context

Stage 2: Fault Alert

Converging DTC codes flag a vehicle scheduled for a high-risk route tomorrow

Unit F-2847 has triggered three independent fault signals over the past 48 hours — a coolant temperature deviation, active brake wear below the service threshold, and a recurring engine misfire pattern. The vehicle is assigned to a 220-mile interstate delivery run departing at 6:00 AM tomorrow. The maintenance decision has to happen now, before the vehicle leaves the yard.

Fleet Health Alert

F-28

Unit F-2847 — 2023 Commercial Cargo Van

87,400 miles | Assigned: Route 14 — Interstate delivery | Departs 6:00 AM tomorrow

MAINTENANCE REQUIRED

Alert typepredictive_fault_convergence

Active DTC codesP0128 (coolant temp) + P0300 (engine misfire)

Brake pad wear12% remaining — below 20% service threshold

Coolant temp deviation+17°F above baseline over last 3 operating hours

Misfire events4 detected in last 60 miles — escalating pattern

Last service6,200 miles ago — overdue by 1,200 miles

Scheduled route risk220 miles — 85-mile stretch with no service coverage

Why This Moment Matters

Three independent fault signals are converging on one vehicle — a vehicle assigned to a long-distance interstate run with an 85-mile gap in service coverage. A breakdown on that route means roadside assistance, driver downtime, missed delivery SLAs, and emergency repair costs that are 4–6x higher than proactive service.

This is where Redis becomes much more than cache. The value is not telemetry speed. The value is assembling the full Vehicle 360 — live fault state, maintenance history, route risk, and parts availability — fast enough to make the right decision before the vehicle leaves the yard.

Stage 3: Ingest

Vehicle, maintenance, and fleet systems flow into Redis

RDI synchronizes vehicle state, maintenance records, and fleet assignment data. Redis Feature Form serves the online predictive features used by the maintenance decision stack. Redis becomes the live working set for fleet health decisions without waiting for slow cross-system joins.

Source Systems → Redis

TELE

Vehicle Telematics / OBD-II

DTC codes, sensor readings, engine metrics, idle time, EV battery state

→

FMS

Fleet Management System

Vehicle assignments, routes, driver schedules, utilization history

→

MAINT

Maintenance Records

Service history, parts replaced, mileage logs, FSA history, warranty status

→

PARTS

Parts & Service Network

Dealer inventory, service bay availability, mobile service dispatch capacity

→

KFK

Kafka / Telemetry Stream

Real-time fault events, harsh acceleration, fuel, speed, and driver behavior feeds

→

Pipeline Status

Vehicle state syncSub-second

Fault event feedStreaming

Maintenance record syncContinuous

Redis Feature Form parity100%

Route assignment pathServed from Redis

Decision modePredictive inline

Stage 4: Context

Redis assembles the live Vehicle 360

The maintenance decision needs fault state, service history, route risk, and parts availability in the same path. That is operational vehicle context — not a DTC code on a dashboard waiting for someone to notice it.

Redis RAMRedis FlexRedis Context Retriever

Durable Vehicle Context

Vehicle mileage87,400 miles — approaching high-wear threshold for this platform

Last service interval6,200 miles ago — overdue by 1,200 miles

Coolant system historyTwo prior coolant warnings logged in past 90 days

Brake service historyRear pads replaced at 72,000 miles — 15,400 miles ago

Misfire FSA patternP0300 code appeared twice in prior 30 days — escalating

Cohort comparison3 of 8 same-model units flagged for proactive service this quarter

Live Fault Context

Active DTC codesP0128 + P0300 — both open and escalating

Scheduled route risk220 miles — 85-mile stretch with no service coverage

Load and terrainFull payload delivery — elevated brake and engine demand

Parts availabilityCoolant thermostat + brake pads in stock at 7-mile service center

Service bay availabilityOpen slot at 7:00 AM today — 1 hour before route window

Current action neededProactive service before departure — not monitor and wait

Context signal: Redis Context Retriever assembles the Vehicle 360 — asset state, maintenance history, and operational context — so the decision engine has exactly what it needs. Converging DTC codes, overdue service interval, route risk, and available service capacity all point to one defensible action.

Stage 5: Feature Serving

Predictive fleet features hydrate in milliseconds

Redis Feature Form serves the online features that support predictive maintenance decisioning: component failure probability, brake urgency, misfire escalation risk, route breakdown exposure, parts availability, and deferral cost — all from vehicle telematics patterns and service history.

coolant_failure_probability

Likelihood of coolant system failure within 500 operating miles given current DTC pattern

0.870.4 ms

brake_wear_urgency

Remaining brake life as a function of scheduled route load and terrain demand

0.940.3 ms

misfire_escalation_risk

Probability of P0300 pattern progressing to drivability failure under load

0.790.5 ms

route_breakdown_exposure

Combined fault severity and route risk score for the scheduled assignment

0.910.3 ms

parts_availability_score

Confidence that required parts are in stock at the nearest accessible service location

0.960.2 ms

deferral_cost_multiplier

Estimated cost ratio of reactive roadside repair versus proactive scheduled service

4.8x0.4 ms

Feature Serving Performance

Features Hydrated

124

P99 Lookup

2.3 ms

Train / Serve Parity

100%

Stage 6: Decision

Proactive service, route reassignment, monitor, or dispatch — one action wins

The maintenance decision engine evaluates the available actions using the unified Vehicle 360. This is not a DTC code on a dashboard. It is an action decision grounded in live fault context, route risk, and service availability assembled by Redis.

#1 Winning Action

PROACTIVE SERVICE

Schedule pre-departure maintenance at 7:00 AM service bay

Parts available, bay open, vehicle serviced before route window — zero-downtime outcome

Brake wear, coolant fault, and misfire pattern all converge. Service capacity exists. This is the only action that eliminates route risk entirely.

Action confidence0.94

#2 Escalate

ROUTE REASSIGNMENT

Swap to Unit F-2891 — healthy vehicle, same route capacity

Valid secondary if service bay were unavailable or parts out of stock

Protects the delivery SLA but defers the fault — vehicle still needs service before the next assignment.

Action confidence0.71

Suppressed

MONITOR & ALERT

Flag for next scheduled service

Suppressed — three converging faults with a high-risk route makes deferral indefensible

This is exactly the action a fragmented maintenance stack is more likely to choose. The alert exists but nobody acts on it before departure.

Action confidence0.08

Stage 7: Business Impact

The value is downtime prevented before the vehicle leaves the yard

Proactive maintenance decisions driven by live vehicle context reduce unplanned breakdowns, protect delivery SLAs, cut emergency repair costs, and keep total fleet utilization high — without requiring a larger fleet to absorb the slack.

Operational Value

Breakdown preventionFault caught before costly roadside failure

Emergency repair costProactive service vs. roadside repair — 4–6x cost difference

Delivery SLA protectionRoute covered by healthy unit — no missed commitments

Driver safetyHigh-risk route not assigned to a vehicle with active fault codes

Fleet utilizationVehicle returned to service faster with planned repair than reactive recovery

Per-Event Outcome

reactive

breakdown on route
roadside + tow + delay

→

proactive

serviced pre-departure
route runs on time

Stage 8: Outcome

Same fleet. Different maintenance quality.

Without Redis, the fleet ops team sees a DTC alert on a dashboard — one of dozens — and the vehicle departs on schedule. With Redis, the Vehicle 360 is assembled in real time so the right action is taken before the vehicle leaves the yard.

Fragmented Fleet Ops

FLEETOPS

F-28

Alert: Unit F-2847

Current state

DTC logged

Alert visible — no action taken

reactive

repair

4–6x

cost

missed

delivery

Redis-Powered Fleet Ops

FLEETOPS

F-28

Alert: Unit F-2847

Winning action

Proactive Service

Vehicle 360 assembled in Redis

Why this wins

Brake wear + coolant fault + misfire + route risk converge

Parts available, bay open at 7 AM. Vehicle serviced before departure — delivery runs on time.

94%

confidence

<15ms

decision

zero

downtime

Stage 9: Architecture Recap

Redis becomes mission-critical decision infrastructure for fleet uptime

Telematics, fleet management, maintenance records, and parts systems stay in place. Redis becomes the operational context layer that makes those systems act together in the live maintenance decision path.

Data Sources

Vehicle Telematics / OBD-II

DTC codes, sensor readings, engine health metrics, EV battery and range state

Fleet Management System

Vehicle assignments, routes, driver schedules, utilization history

DTC + FSA Database

Fault code definitions, field service action history, severity index

Parts & Service Network

Dealer inventory, service bay availability, mobile service capacity

Kafka / Telemetry Stream

Real-time fault events, idle, speed, harsh acceleration, and fuel feeds

Ingest Layer

RDI

Syncs vehicle, maintenance, and fleet assignment state

Redis Feature Form

Serves predictive failure, service urgency, and route risk features online

Unified Context Layer

Redis RAM

Hot fault path, active DTC alerts, live vehicle health state

Redis Flex

Warm maintenance history, failure patterns, and vehicle embeddings

Feature Store

Failure probability, component wear, utilization, and route risk

Redis Context Retriever

Assembles the Vehicle 360 — asset state, maintenance history, and operational context — and exposes it as structured MCP tools for the decision engine

Decision Engine

Fault Severity Rules

DTC threshold scoring, FSA priority, and safety-critical flag logic

Predictive Failure Scorer

ML-based component failure probability from telematics patterns

Service Arbitration

Route risk, parts availability, and schedule constraint resolution

Redis Search

Failure typology matching across historical fleet fault patterns

Output Actions

Proactive Service

Schedule pre-route maintenance before vehicle departs

Route Reassignment

Swap vehicle assignment to available healthy unit

Monitor & Alert

Flag for next scheduled service with active telemetry watch

Emergency Dispatch

Mobile service or roadside assistance with full fault context

Decision Target

<15 ms

Primary Goal

Uptime + fleet utilization

Redis Role

Real-time vehicle context