The Connected Vehicle Data Challenge
Modern vehicles are sophisticated computing platforms generating massive amounts of data. A single car produces nearly 25GB of data per hour—telematics, GPS coordinates, sensor signals, infotainment activity, diagnostic logs. Scale that across millions of vehicles, and the engineering challenges become profound.
But volume is only part of the problem. The real complexity lies in how this data flows, transforms, and synchronizes across a fragmented ecosystem.
The Technical Challenges
1. Data Model Fragmentation
Each actor in the connected vehicle ecosystem—OEMs, charging networks, infrastructure providers, fleet operators—defines vehicle data differently. Without standardization, integrating data across systems requires constant translation layers and custom mappings.
COVESA's Vehicle Signal Specification (VSS) addresses this by providing a standardized vocabulary for vehicle signals. But implementing VSS across a distributed system introduces new challenges: how do you enforce schema consistency while maintaining flexibility? How do you evolve data models without breaking existing systems?
2. Edge-to-Cloud Synchronization
Vehicles operate in bandwidth-constrained, intermittently-connected environments. Data must flow reliably from vehicle ECUs to cloud systems while handling:
- Connectivity Interruptions: Vehicles lose connectivity regularly. Systems must queue data locally and sync when connectivity returns
- Conflict Resolution: When multiple systems update the same data, conflicts must be resolved deterministically
- Bandwidth Optimization: Sending 25GB/hour per vehicle to the cloud is impractical. Delta sync (only sending changes) is essential
- Latency Sensitivity: Some data (safety-critical telemetry) requires near-real-time delivery; other data (historical logs) can be batched
3. Multi-Model Data Requirements
Connected vehicles generate diverse data types that don't fit neatly into a single model:
- Time-Series Data: Sensor readings, performance metrics, diagnostic logs
- Geospatial Data: Vehicle location, route history, proximity queries
- Transactional Data: Charging sessions, payment records, user interactions
- Graph Data: Vehicle relationships, infrastructure networks, supply chains
- Vector Data: Embeddings for anomaly detection, predictive maintenance
Traditional relational databases force this diverse data into rigid schemas. Document-oriented approaches provide flexibility but require careful design to maintain consistency.
4. Real-Time Query Performance at Scale
Fleet management, EV charging networks, and autonomous systems require instant answers to complex queries:
- "Find all vehicles within 5km of this charging station"
- "Identify vehicles with anomalous battery degradation patterns"
- "Calculate optimal routes for 10,000 vehicles in real-time"
- "Aggregate telemetry from 1 million vehicles in the last hour"
Volvo Connect processes 65 million daily events from over a million vehicles. SHARE NOW handles 2TB of IoT data per day from 11,000 vehicles across 16 cities. These workloads require databases designed specifically for this scale and query pattern.
Architectural Approaches
Standardized Data Models as Foundation
The first step is adopting industry standards like VSS. But standards alone aren't enough. The architecture must:
- Map VSS concepts directly to storage structures (avoiding translation overhead)
- Support schema versioning for backward compatibility
- Enable gradual migration as data models evolve
- Provide validation at write-time to catch inconsistencies early
Embedded Databases on ECUs
Rather than streaming all data to the cloud, embed lightweight databases directly on vehicle ECUs. This enables:
- Local data persistence during connectivity loss
- Intelligent filtering (only send relevant data to cloud)
- Reduced bandwidth consumption
- Faster local queries for in-vehicle applications
The challenge: keeping embedded and cloud databases synchronized while handling conflicts and maintaining consistency.
Multi-Model Storage Strategy
Use a database that natively supports multiple data models:
- Time-series collections for sensor data (optimized for sequential writes and time-range queries)
- Document collections for transactional data (flexible schema, rich queries)
- Geospatial indexes for location-based queries (sub-millisecond performance)
- Vector support for AI/ML workloads (embeddings for anomaly detection)
This eliminates data movement between systems and simplifies the application layer.
Horizontal Scalability by Design
As vehicle fleets grow from thousands to millions, the database must scale horizontally:
- Sharding strategies that distribute data by vehicle ID, region, or time
- Transparent sharding that doesn't require application changes
- Cross-shard queries that aggregate data efficiently
- Replication for high availability and disaster recovery
Edge-to-Cloud Synchronization Protocol
Design a protocol that handles the realities of vehicle networks:
- Delta Sync: Only transmit changes, not full documents
- Compression: Reduce bandwidth consumption
- Conflict Resolution: Deterministic rules for concurrent updates
- Batching: Group small updates into larger payloads
- Prioritization: Send critical data immediately, batch non-critical data
Real-World Considerations
Compliance and Audit Trails
Connected vehicles operate in regulated environments. The data platform must:
- Log all data access and modifications
- Support compliance frameworks (SOC 2, ISO 27001, automotive-specific standards)
- Enable audit queries across millions of records
- Maintain data integrity for legal proceedings
Predictive Maintenance and AI
Modern fleet management uses AI to predict failures before they occur. This requires:
- Aggregating historical data from millions of vehicles
- Computing features (statistical summaries, time-series patterns)
- Training models on this data
- Scoring new data in real-time
The database must support both batch analytics and real-time scoring without data movement.
Multi-Cloud and Hybrid Deployments
Automotive companies often operate across multiple cloud providers and on-premises infrastructure. The platform must:
- Synchronize data across cloud boundaries
- Support hybrid edge-cloud deployments
- Maintain consistency across regions
- Enable disaster recovery across providers
Lessons Learned
1. Schema Flexibility Doesn't Mean Schema Chaos
Document databases provide flexibility, but connected vehicle systems benefit from enforced schemas. Use schema validation to catch errors early while maintaining the ability to evolve schemas over time.
2. Standardization Enables Scale
COVESA's VSS isn't just a data model—it's a foundation for interoperability. Systems that adopt standards early can integrate with partners more easily and scale faster.
3. Edge Computing is Essential
Trying to stream all vehicle data to the cloud is impractical. Embed databases on ECUs, process data locally, and sync intelligently. This reduces bandwidth, improves latency, and enables offline functionality.
4. Multi-Model Databases Reduce Complexity
Rather than maintaining separate databases for time-series, geospatial, and transactional data, use a platform that handles all of these natively. This simplifies operations and eliminates data movement.
5. Real-Time Performance Requires Purpose-Built Infrastructure
Generic databases struggle with connected vehicle workloads. Purpose-built infrastructure—with native time-series support, geospatial indexing, and horizontal scalability—is essential.
Looking Ahead
The connected vehicle ecosystem is evolving rapidly. Future challenges include:
- Agentic AI: Autonomous agents making decisions based on vehicle data
- OTA Updates: Safely updating vehicle software and data schemas over-the-air
- Privacy-Preserving Analytics: Extracting insights from vehicle data without exposing individual user information
- Interoperability at Scale: Seamlessly integrating data across OEMs, charging networks, and infrastructure providers
These challenges require databases designed from the ground up for connected vehicle workloads.
Samuel M.K
Founder & CTO
CredVault
April 16, 2026