Definition
Data Synchronization (Data Sync) refers to the process of keeping data consistent across multiple data sources (devices, platforms, regions). In the context of the Internet of Things, data synchronization includes: Device and Platform — Device data collection reported to the platform, platform configuration and firmware sent to the device; Between Multiple Platforms — Charging operation platforms and eMSPs synchronize charging sessions and CDRs through OCPI; Across Multiple Regions — When deployed cross-border, data from the Device Management Platform in each region is synchronized to the central platform. Synchronization strategies include full synchronization, incremental synchronization, differential transmission , combined with differential calibration to ensure consistency.
Data synchronization is the foundation for cross-border business collaboration and requires consideration of network latency, conflict resolution, and eventual consistency.
Technical Features and Architecture
Features of data synchronization technology include: Synchronization Direction — one-way (device→platform), two-way (device↔platform), multi-way (multi-platform mutual sync); Synchronization Strategy — scheduled full sync, real-time incremental sync, change-triggered ; Conflict Resolution — last-write-wins, version vectors, CRDTs, etc.; Transmission Optimization — data compression , differential transmission to reduce the amount of synchronized data. In differential calibration scenarios, full synchronization is used to calibrate the baseline, while daily operations use incremental synchronization.
In cross-border scenarios, data synchronization needs to consider time zones, unstable networks, and compliance (restrictions on cross-border data transmission). A hierarchical architecture of regional centers + central aggregation can be adopted.
Application in IoT and cross-border scenarios
Applications of data synchronization in IoT and cross-border scenarios include: Charging interoperability — OCPI enables the synchronization of charging sessions and CDRs between CPOs and eMSPs; Device management — Device status and configuration are synchronized between devices and Device Management Platform ; Cross-border logistics — Orders and Location are synchronized across multiple platforms and regions; Cross-border operations — Data from cross-border deployed EV charging stations and logistics devices are synchronized to the headquarters platform. SimtoLink provides IoT devices with connectivity covering over 200 countries, supporting reliable data synchronization in cross-border scenarios.
Consistency Limits When Several Copies Must Match
Full, Incremental, and Eventually Consistent
Data synchronization means the device shadow, the regional platform, and the central store reach an explainable agreement on the same version of the same object. It does not require every byte to match at every instant. A full sync covers the baseline with a complete snapshot. It fits first join, reconnect after a long offline period, and the moment a version no longer matches. An incremental sync moves only changed fields. Daily traffic is smaller, but the receiver must already hold the same baseline. Eventual consistency allows a cross-border path to show an old value first, as long as a version number or a vector clock can converge later. Strong consistency requires a majority of copies to confirm before a write succeeds. Control commands and billing state are closer to that requirement. A telemetry curve usually cannot afford that round-trip delay.
Cross-Border Time Zones and Catch-Up Windows
The conditions of use are these. Each record carries a stable object id, a monotonic version or sequence number, and a clear time source. The conflict policy is written in advance: last write wins, merge by field, or hold the disagreement for a person. While offline, the device can queue locally and, after reconnect, catch up by sequence number, rather than reporting only the "current value." A cross-border time zone cannot decide order from a local wall clock. Use a timestamp with a time zone, or use the sequence number alone. When a regional center collects local devices first and then aggregates to headquarters, mark which layer this record has reached. Otherwise headquarters treats a hole that has not yet been aggregated as a device that never reported. If the sync path crosses a roaming network, rate-limit the batch catch-up. A device that comes online must not fill the link and push real-time alerts aside.
Treating an Out-of-Order Packet as the Latest Value
A common mistake overwrites state in arrival order. During a roaming handover a later-sent old packet can arrive last. Without a version comparison, the platform writes a recovered device back to fault. Another mistake runs bidirectional sync with no primary. Configuration pushed by the platform and configuration changed on the device overwrite each other, and neither side knows which copy is valid. Incremental sync without differential calibration lets the baseline drift, and every later increment stacks on the wrong start. Delete and "field not present" must be distinguished, or the peer treats a missing field as a clear to zero. Clock jumps, daylight saving, and a failed operator time-by-SMS all make time-ordered sync step backward. A successful sync task means only that the message was accepted. It does not mean the business meaning already matches. Critical state should also be read back.
Regional aggregation should also mark the state "arrived locally, not yet at the center." Otherwise the headquarters view draws a hole still in transit as a silent device. The catch-up queue needs a cap and a drop policy. Keep alerts and configuration receipts first, and drop stale high-rate samples, rather than letting first-in-first-out push out the newest fault. Delete sync needs a tombstone. Do not guess a delete because a field is absent from the peer's next full snapshot. If both sides change the same parameter and neither is primary, the side that arrives later wins, and staff on site see a threshold they just set covered by the old value. Acceptance should manufacture out-of-order, duplicate, and partial packets, and confirm the state machine stops in front of the gap. Watching sync succeed only inside a lab with a good network says nothing about a roaming handover.